BLIND SEPARATION USING ABSOLUTE MOMENTS BASED ADAPTIVE ESTIMATING FUNCTION
不顾分歧或不赞成而追求爱情的英语作文
不顾分歧或不赞成而追求爱情的英语作文全文共3篇示例，供读者参考篇1Defying Disapproval: The Relentless Pursuit of True LoveThey say love knows no bounds, that it transcends all barriers and conquers even the tallest of obstacles. But what happens when those obstacles are the people closest to you - your own family and friends? What if the very foundations you were raised upon crack and crumble at the prospect of your heart's desires? For me, this wasn't just a hypothetical dilemma, but a harsh reality I had to confront head-on.From the moment Sarah and I first locked eyes in freshman English class, I knew there was something special about her. Her radiant smile, infectious laughter, and compassionate soul captivated me in a way I had never experienced before. We became fast friends, bonding over our shared love of literature and dream of one day becoming writers.As the months passed, our friendship blossomed into something deeper, something richer than I could have ever imagined. The lingering glances, the electric touch of our handsbrushing against each other, the countless nights spent analyzing every nuance of our favorite novels – it all culminated in an eruption of passion and vulnerability unlike anything I had felt in my 18 years on this earth.But our love, as pure and true as it was, did not come without its fair share of complications. Sarah came from a devoutly religious family, one that subscribed to traditional gender roles and frowned upon any deviation from the heteronormative path laid out before her. My own parents, children of the 60s with hippie ideals, harbored their own reservations about my entanglement with someone from such a diametrically opposed background.The first signs of disapproval came in hushed whispers and loaded glances whenever Sarah and I were together. The judgmental stares from her parents whenever I picked her up for our dates. The thinly-veiled jabs from my mother about "seeing the world" before "settling down." It was as if our families could sense the depths of our connection, and were determined to sever those bonds before they became unbreakable.As the tension escalated, so too did the confrontations. Sarah's father, a fire-and-brimstone preacher, sat us down for an excruciating three-hour lecture on the sanctity of traditionalmarriage and the "sinful" nature of our relationship. My parents, ever the free spirits, expressed concerns about the stringent dogma that governed Sarah's upbringing, fearing it would stifle her independence and personal growth.Through it all, Sarah and I remained resolute, our love a steadfast anchor amidst the tempestuous seas of judgment and criticism. We knew, deep down, that what we shared was something incredibly rare and precious – a bond that transcended the rigid confines of societal expectations and familial mandates.Still, the constant barrage of negativity took its toll. There were nights when the weight of it all became too much to bear, when the thought of succumbing to the pressures seemed like the easier path to take. But then I would look into Sarah's eyes, those endless pools of warmth and understanding, and find the strength to carry on.Our families' opposition only seemed to intensify as our relationship deepened, as if they could sense the permanence of our commitment. Threats of disownment and estrangement were brandished like weapons, desperate attempts to drive an immutable wedge between us.Yet, through it all, we persevered. We found solace and support in one another, creating our own little oasis of love amidst the raging storms of disapproval. Our bond was forged in the fires of adversity, tempered by the challenges that so many said would break us.And now, as I stand here today, my hand intertwined with the woman who has become my entire world, I can look back on that tumultuous journey with a sense of pride and accomplishment. Our love prevailed, not in spite of the difficulties we faced, but because of them.To those who may find themselves in a similar situation, caught between the intense gravitational pulls of societal expectations and the undeniable yearning of their hearts, I offer this advice: have courage. Have the fortitude to fight for what you believe in, for the person who sets your soul ablaze with a love so pure and transcendent that it defies all attempts to extinguish it.Love is not a whimsical fancy, but a force of nature –powerful, unrelenting, and destined to overcome any obstacle in its path. Those who would stand in its way do so at their own peril, for true love cannot be contained or subdued by the narrow-minded conventions of others.So embrace that love, nurture it, and guard it fiercely against those who fail to recognize its majesty. For in the end, when you gaze into the eyes of your soulmate and know that you both prevailed against all odds, there is no greater victory, no sweeter redemption than that.Our story may have begun with disapproval and discord, but it will end in eternal harmony, two souls intertwined for a lifetime of understanding, acceptance, and unconditional love. And that, my friends, is a triumph worth fighting for.篇2Defying All Odds: A Story of Love Conquering DifferencesThey say love is blind, but I've come to realize it's more than that - love can make you defiantly bold in the face of opposition. My relationship with Jake was the kind that made people raise their eyebrows and voice their concerns from the start. We came from completely different worlds - I was the academic overachiever headed to an Ivy League university, while Jake was the high school dropout working odd jobs to make ends meet. On paper, we made no sense at all. But the heart wants what it wants, and despite the incessant voices warning me against thismismatched romance, I couldn't deny the magnetic pull I felt toward Jake.I still remember the day we met like it was yesterday. I was at the local diner cramming for my calculus final when this rowdy bunch of guys walked in, fresh off a construction job by the looks of their dirt-stained clothes and boisterous laughter. Jake stood out immediately - not because of his disheveled appearance, but because of the warm, infectious smile that lit up his face. Our eyes met for the briefest moment, and I felt an instant connection, an unspoken mutual understanding that transcended our surface-level differences.From that fateful encounter, an unlikely friendship blossomed. Jake would come into the diner on his lunch breaks, and we'd spend hours talking about everything from classic literature to the meaning of life. He had this profound wisdom about him that belied his rough exterior, and I found myself captivated by his unique perspective. Before long, our conversations turned to lingering glances, and those glances ignited a spark that couldn't be ignored.When we finally mustered the courage to make our feelings known, the world around us seemingly crumbled. My friends were appalled that I, the embodiment of academic excellence,would entertain the idea of dating a high school dropout with no clear path. Jake's buddies ribbed him endlessly about "slumming it" with a rich kid who couldn't possibly understand his struggles. Our families were perhaps the loudest voices of dissent - my parents threatened to disown me if I pursued this "unbecoming" relationship, while Jake's blue-collar relatives warned him about the hazards of dating someone from the "ivory tower."But in the eye of that judgmental storm, our connection only grew stronger. We found solace in each other's embrace, a safe haven from the cacophony of unsolicited opinions and societal expectations. Jake showed me a world beyond the confines of my privileged upbringing, teaching me to appreciate the simple joys in life and the dignity of hard work. In turn, I introduced him to the wonders of higher learning, encouraging his natural curiosity and thirst for knowledge.Together, we embarked on a journey of mutual growth and self-discovery, defying the odds and the naysayers at every turn. When I received my acceptance letter to an elite university across the country, it was Jake who inspired me to take the leap, leaving the comforts of my hometown behind. He promised to follow me wherever life took us, and true to his word, he worked tirelessly to save up enough money to join me after my first year.Those early days of our long-distance relationship were grueling. The skeptics had a field day, proclaiming that our love would wither under the strain of separation and diverging paths. But we were determined to prove them wrong. We wrote heartfelt letters documenting our daily lives, sharing our hopes and fears with an intimacy that transcended physical proximity. When Jake finally made his way to me, brick by calloused brick, we built a life together that shattered every preconceived notion about our unlikely pairing.It wasn't always easy, of course. We had our fair share of arguments, often stemming from the fundamental differences in our backgrounds and perspectives. But we learned to embrace those differences, using them as opportunities for growth and compromise rather than points of contention. We carved out a unique dynamic that defied traditional labels and roles, allowing each other to shine in our respective strengths while lifting each other up in our weaknesses.As the years passed, the voices of doubt and disapproval faded into distant echoes, drowned out by the resounding triumph of our enduring love. We watched in awe as our families, once vehemently opposed to our union, slowly came around to accept and even embrace our unconventional partnership. Ourfriends, who had once judged us harshly, became our biggest cheerleaders, inspired by the depth of our commitment and the resilience of our bond.And through it all, Jake and I remained steadfast in our belief that love, in its purest form, transcends all barriers and overcomes all obstacles. We became living embodiments of the idea that true love knows no boundaries, be they socioeconomic, cultural, or otherwise. Our story was a testament to the power of following one's heart, even when the world seems determined to tear it apart.As I reflect on our journey, I am filled with an overwhelming sense of gratitude for the lessons we've learned along the way. Love, we discovered, is not a fairytale or a fleeting infatuation –it's a choice, a daily commitment to weather the storms and embrace the challenges that life throws our way. It's a tapestry woven from the threads of patience, compromise, and unwavering support, ever-evolving and growing stronger with each passing year.To those who find themselves in the throes of a love that defies societal norms or faces opposition from those around them, I offer this advice: have courage, and follow your heart. The road may be long and arduous, but the reward of a love thatovercomes all obstacles is worth every ounce of struggle. Embrace your differences, for they are the fertile soil from which the most beautiful connections bloom. And never, ever, let the voices of doubt and judgment drown out the whispers of your soul – for in those whispers lie the key to a love that transcends all boundaries.In the end, Jake and I learned that love is not blind, but rather, it is the guiding light that illuminates the path forward, even in the face of seemingly insurmountable odds. Our story is a testament to the indomitable power of love, a living embodiment of the truth that when two hearts are united in their resolve, no force on earth can tear them apart.篇3The Choice of the HeartThey say love is blind, but I've found it has perfect vision when it comes to seeing the truth of who someone really is on the inside. The kind of love I'm talking about isn't just physical attraction or infatuation - it's the bone-deep knowing thatyou've found the other half of your soul.I felt that profound connection from the first moment I looked into Zayn's eyes. We came from completely differentworlds, but it didn't matter. His family was strict Muslim and had very traditional expectations when it came to courtship and marriage within their faith. My parents were progressive atheist intellectuals who looked down on religious traditions as antiquated and unenlightened.From the outside, it seemed Zayn and I were totally incompatible. But we both had a free spirit and rebellious streak that bonded us together. We enjoyed having deep conversations about philosophy, trading perspectives from our vastly diverse upbringings. Even when we didn't agree, we listened to each other with openness and curiosity.Slowly, meeting up after classes turned into lingering study sessions at the library. Casual talks escalated into breathless romantic tension. Zayn made me feel understood and appreciated for my most authentic self - insecurities, quirks, and all. In his presence, I didn't have to censor my unconventional thoughts and opinions to fit anyone's standards but my own.The first time he took my hand, sparks visibly flew between us. In that static-charged moment, I knew I was falling hard. I'd never been risk-averse, but letting myself love Zayn so completely felt dangerous - even more so because of theinevitable conflicts it would cause with our families' clashing beliefs and traditions.Sneaking around was simultaneously thrilling and heartbreaking. We cherished our stolen moments together - furtive glances, whispered endearments, secret coffee dates. Every brush of his hand set my skin alight with yearning. Yet we could never fully relax into our feelings as judgment loomed from both our households.My parents hoped I'd pick a "suitable" partner to uphold our secular values -eitherravant academic or a modern free-thinker. They worried I was tarnishing my reputation and throwing away future prospects by consorting with a religious conservative like Zayn.His parents anticipated an arranged marriage with a "proper" Muslim woman who would sacrifice her own identity to become a traditional wife and homemaker within their tight-knit community. If Zayn followed his heart's forbidden desire for me, he'd face disownment and disgrace.As our relationship escalated despite all objections, tensions mounted within our families and friend circles. Cutting insults and cruel assumptions got flung from both sides, each groupdamning the other as brainwashed or oppressive based on parochial stereotypes.In the midst of that cross-cultural war zone, Zayn and I were a secret oasis of acceptance. We didn't try to change or convert one another - we honored each other's fundamental identities. I'd never been Muslim or stepped foot in a mosque, but I stood in awe as Zayn described finding rapturous beauty and divine peace in his spiritual practices.I shared how my atheist beliefs revolved around individuality, human potential, and breaking the shackles of dogma to embrace freedom of thought. Zayn admired my boldness for questioning norms and daring to walk my own path.In those raw, vulnerable moments, our souls became inextricably entwined. Our love transcended ethnicity, ideology, and all the other superficial boundaries humanity erects between itself.Still, we couldn't live in our insulated bubble forever. Keeping our relationship under wraps grew increasingly difficult - and dishonest. How could something so pure and profound between two people be considered shameful? Why did we have to hide how we felt from those closest to us?After agonizing for months, Zayn and I decided to go public and face the consequences together. His parents reacted even more devastatingly than I imagined, launching into a tornado of anger, threats, and explosive emotion. My mom and dad simply went icy cold with bitter disappointment.With our loved ones refusing to acceptance our relationship, we reached a make-or-break crossroads. Did we sacrifice our truest selves to conform to others' limited expectations and visions for us? Or did we take the road less traveled and fight for a love that defied all conventions and odds?The old me might have cowered away from such profound disruption and hurt. But Zayn's devotion gave me a newfound resilience, empowering me to stand up for the desires of my heart without apology.In the end, we chose each other.Zayn was disowned by his devout Muslim family. My secular parents eradicated me just as harshly, burning ties over my supposed betrayal of their principles. We became cultural refugees - rejected by our respective tribes, yet bonded in our mutually outcast status.Looking back years later, I have no regrets despite how excruciatingly painful and lonely that period was. If I'd let fear and misperceptions dictated my path, I never would have embarked on the most fulfilling relationship and beautiful life partnership I could。
声源定位技术文献综述和英文参考文献
声源定位技术文献综述和英文参考文献声源定位在各个领域都有着广泛的应用，早在20世纪七八十年代，声源定位系统就开始被广泛地研究，尤其是基于传感器阵列的方法。
它的应用使得电话会议、视频会议、可视电话等系统中摄像头和传声器能够对准正在说话的人。
30471声源定位技术在经过几十年的发展后，其检测技术已经有了极大程度的发展和提高。
由最早的基于碳粒子或冷凝器来接收声信号的模式的普通声波检测技术发展到如今基于电路集成化与电子信息化结合的声源检测技术。
现代的声源定位现代技术测量过程简化了，而检测精度提高了。
论文网国外的声波检测技术已经在坦克和武装直升机上得到了广泛的应用，而在这方面，传感器技术、探测技术、微电子技术、信号处理技术以及人工智能技术的飞速发展，均为声源探测技术用于直升机等军事目标的定位、跟踪和识别开辟了新的应用前景，使声源探测技术成为一种重要的军事侦察手段和防空作战中反电子干扰和反低空突防的一种有效途径。
当然国内在这方面的研究也是逐步与国际接轨。
近年来，具有广阔的应用前景和实际意义的声源定位技术已成为新的研究热点，不仅仅是在军事上，许多国际著名公司和研究机构已经在声源定位技术研究与应用上开始了新的角力，许多产品已进入实际应用阶段。
并且已经显示出巨大的优势和市场潜力。
参考文献[1] Oyilmaz,S.Rickard. Blind Separation of Speech Mixtures via Time-Frequency Masking[J]. IEEE Transactions on Signal Processing, XX, 52(7):1830-1847.源自[2] H. Sawada, S. Araki, R. Mukai, S. Makino. Blind extraction of dominant target sources using ICA and time-frequency masking[J]. IEEE Transactions on Audio, Speech, and Language Processing , XX, 14 (6): 2165–2173.[3] M.Swartling,N.Grbic´, I.Claesson. Direction of arrival estimation for multiple speakers using time-frequency orthogonal signal separation[C]. Proceedings of IEEE International Conference on acoustic, Speech and Signal Processing, XX. 833–836.[4] M. S. Brand stein, J.E. Adcock, H.F. Silverman.A closed-form location estimator for use with room environment microphone arrays[J]. IEEE Transactions on Speech and Audio Processing, 1997, 5 (1): 45–50.[5] M. Swartling, M. Nilsson, N.Grbic. Distinguishing true and false source locations when localizing multiple concurrent speech sources[C]. Proceedings of IEEE Sensor Array and Multichannel Signal ProcessingWorkshop, XX. 361–364.[6] E. Di Claudio, R. Parisi, G. Orlandi. Multi-source localization in reverberant environments by ROOT-MUSIC and clustering[C]. Proceedings of IEEE International Conference on Acoustic, Speech and Signal Processing, XX. 921–924.[7] T. Nishiura, T. Yamada, S. Nakamura, K. Shikano. Localization of multiple sound sources based on a CSP analysis with a microphone array[C]. Proceedings of IEEE International Conference on Acoustic, Speech and Signal Processing, XX. 1053–1056.[8] R. Balan, J. Rosca, S. Rickard, J. ORuanaidh. The influence of windowing of time delay estimates[C]. Proceedings of Conference on Information Sciences and Systems, XX. 15–17.[9] S. Shifman, A. Bhomra, S. Smiley, et al. A whole genome association study of neuroticism using DNA pooling[J]. Molecular Psychiatry, XX, 13(3): 302–312.[10] S. Rickard, R. Balan, J. Rosca, Real-time time-frequency based blind source separation[C]. Proceedings of International Workshop on Independent Component Analysis and Blind Signal Separation, XX. 651–656.源自[11] K. Yiu, N. Grbic, S. Nordholm, et al. Multi-criteria design of oversampled uniform DFT filterbanks[J]. IEEE Signal Processing Letters, XX, 11(6): 541–544.[12] E. Vincent. Complex nonconvex lp nom minimization for underdetermined source separation[C]. Proc. ICA, XX. 430-437[13] C. Knapp , G. Carter. The generalized correlation method for estimation of time delay[J]. IEEE Trans. Acoust., Speech, Signal Process, 1987, 24(4): 320–327.[14] T. W. Anderson. Asymptotic theory for principal component analysis[J]. Ann. Math. Statist., XX, 34(1): 122–148.[15] D. Campbell, K. Palomäki, G. Brown. A matlab simulation of shoebox room acoustics for use inresearch and teaching[J]. Comput. Inf. Syst. J., XX, 9(3): 48–51[16] J. Huang, N. Ohnishi, N. Sugie. A biomimetic system for localization and separation of multiple sound sources[J]. IEEE Trans. Instrum.Meas., 1995(44): 733–738.[17] B. Berdugo, J. Rosenhouse, H. Azhari. Speakers’direction finding using estimated time delays in the frequency domain[J]. Signal Processing, XX, 82(1): 19–30.[18] S. T. Roweis. One microphone source separation[J]. Neural Inform.Process. Syst., 793–799.[19] J.-K. Lin, D. G. Grier, J. D. Cowan. Feature extraction approachto blind source separation[C]. Proc. IEEE Workshop Neural NetworksSignal Process, 1997. 398–405.[20] M. Van Hulle. Clustering approach to square and nonsquare blind source separation[C]. IEEE Workshop Neural Networks Signal Processing. 1999. 315–323. :。
BLIND SOURCE SEPARATION
∗ ∗ Cq p [si ] = Cum[si (n), . . . , si (n), si (n), . . . , si (n)], p terms q terms def
do not depend on n; for definitions of cumulants, refer to [7] and references therein. H3. At most one source has a zero marginal cumulant of order r. ˘ (z ) = F ˘ (z )H ˘ (z ), satisH4. The global transfer matrix, G H ∗ ˘ ˘ fies the property G(z )G (1/z ) = I where I denotes
,y ∗ ∗ C2 2 [i, j , ℓ] = Cum[yi (n), yi (n) , yj 1 (n − ℓ1 ), yj 2 (n − ℓ2 ) ] (1) where j = (j1 , j2 ) and ℓ = (ℓ1 , ℓ2 ). Also define L a subset of Z2 and J = {1, 2, . . . , N }2 ; unless otherwise specified, L = Z2 . We are now in a position to state the proposition below:
EEG伪影详解和过滤工具的汇总（二）
EEG伪影详解和过滤工具的汇总（二）在《EEG伪影类型详解和过滤工具的汇总(一)》，我们详细介绍了EEG伪影类型和产生原因，这篇文章，我们主要介绍常见脑电伪影的处理技术。
脑电伪影过滤技术(通过数据分析)根据数据分析，处理伪影主要有四种方法：1.脑电伪影剔除第一种方法是对带有伪影的脑电周期进行选择和剔除。
不同的技术定义了一种模式(通常是上述伪影之一)来选择要去除的脑电图epoch。
模式识别方法的范围从脑电图专家的目视检查，到在时域或频域的自动统计(Nolanet al., 2010)。
例如，在ERPs协议中，自己可以定义一个统计阈值，以删除振幅明显更高的试验。
剔除是一种非常昂贵的方法，因为虽然它可以消除几乎所有的伪影，但同时也消除了该epoch的所有有价值的EEG信息。
通常，你会尽可能保留更多的脑电图数据，特别是当记录很短的时候。
2. 过滤这些技术的目标是消除伪影，同时保持尽可能多的EEG图信息。
这种分类包括以下技术：简单的线性滤波器去除某些频段(Panych et al .,1989)；回归方法使用参考信号从EEG中去除EOG或ECG信号(Wallstrom et al ., 2004),自适应滤波器与参考信号(Marque et al ., 2005),维纳滤波器(Sweeney et al ., 2012)或贝叶斯过滤器(Sameni et al ., 2007)。
例如，我们可以使用线性滤波器去除50 Hz或60 Hz的交流电干扰。
这也将消除EEG信息(脑电波)，不过，这种高频通常不是EEG研究的重点。
另一个示例是使用EOG信号作为参考通道，以通过回归或自适应滤波器从受污染的EEG信号中去除这些信息。
回归方法假设记录的脑电图是真实脑电图和伪影(EOG)的结合。
回归滤波器计算在单个EEG通道中存在的参考(EOG)的比例，并将其减去。
3.盲源分离这些是分离技术，试图将脑电图分解成基于不同数学考虑(如正交性或独立性)的信号源的线性组合。
同频信号分离在认知无线电中的应用
同频信号分离在认知无线电中的应用
THE APPLICATION OF THE SEPARATION OF THE SIGNALS WITH SAME FREQUENCY IN THE COGNITIVE RADIO
ABSTRACT
With the rapid development of the wireless communication business, more and more radio spectrum resources are needed, it makes that the current spectrum resources become less and less. Therefore, the study of the cognitive radio technology has great realistic significance.
KEY WORDS: cognitive radio, spectrum sense, energy detection, adaptive filter
ii
河北工业大学硕士学位论文
第一章 绪论
§1-1 课题的研究背景
随着越来越多无线通信业务的发展，频谱资源面临着前所未有的严峻局面。一方面，频谱资源是有 限资源，众多宽带业务的增长，使得现有的频谱资源越来越紧缺；另一方面目前基于固定频谱分配的频 谱资源利用率是很低的[1]。图 1.1 为北京 698-806MHz频谱的使用情况（地点：北京市海淀区，时间： 连续 7 天监测），从图中可以看出，698-750MHz述
同频信号，顾名思义，就是指载频相同的两路信号，但这两路信号可以是不同的调制方式或者是不 同的信号强度，以便用于区分。
在日常的无线电监测中，经常会遇到具有相同载频，但强弱不同的两个信号混合在一起进行传输的 情况，这种干扰严重影响了正常的无线通信业务，是无线电监测工作需要着重解决的问题之一。例如对 卫星链路的非法盗用和恶意干扰就是属于这种情况。再比如，不同的电台是用不同的载波频率来区分， 因此对于同一载波频率的不同电台，接收机往往很难进行区分。当两电台发射的同频信号到达接收点的
欠定盲源分离与信号源估计方法研究
欠定盲源分离与信号源估计方法研究何琪邱晨（长安大学信息工程学院陕西·西安710064）摘要近年来，盲源分离算法研究主要集中在两个方面，混合矩阵估计和源信号个数估计，本文基于理论研究提出了一种盲源分离算法，以语音信号为例，本文采用STFT将语音信号转换到时频域进行分析。
基于现实中很多语音数据通常是高度混叠的信号，所以需要去燥降低信号噪声对混合矩阵和源个数估计的影响。
为了抑制噪声对检测自动源TF点的影响，提出了一种通过使用STFT的主成分分析（PCA）来检测源的自动定位的方法。
另外，基于子空间投影和聚类方法，提出了一种估计混合矩阵的有效方法，使用自动谱聚类方法实现对源个数的估计。
关键词盲源分离信号源估计方法中图分类号：TN911.7文献标识码：A1研究现状盲源分离（BSS）是基于来自传感器阵列或单个传感器的观察到的混合物来恢复基础源信号，而不知道源和混合过程的信息。
对于瞬时欠定传感器阵列的问题，即传感器的数量多于一个但少于源的数量。
带有传感器阵列的BSS比单个传感器的研究更广泛。
这只是因为更多的传感器可以从来源收集更多信息，这有助于分离过程。
BSS问题在音频，雷达，通信，图像处理等领域已经得到广泛应用。
现有的大多数文献已经对高信噪比环境下的语言信号的处理取得了很好的表现。
然而现实中的信号可能被强噪声干扰，这给BSS问题研究带来非常差的实验效果，所以提高算法稳健性势在必行。
此外，BSS算法的有一个问题是未知源数的估计。
在很多研究文献中可以看到对源数估计都是在理论上假设源的个数再使用聚类方法进行聚类，而在实际问题中源个数信号不可用，盲目估计混合信息的源个数会对BSS算法产生影响。
由于短时傅里叶变换（STFT）易于实现且不存在TF域中的交叉项，本文研究语音信号在STFT上的盲源分离算法。
Assa-El-Bey等人提出的STFT-UBSS通过将位于每个自动源TF点的估计STFT值分配给它们对应的源来分离STFT域中的混合源，然后使用已经分配给该源的估计的STFT值通过TF合成来恢复每个源。
基于边缘模糊频谱特征的散焦参数估计方法
基于边缘模糊频谱特征的散焦参数估计方法作者：梁敏朱虹来源：《计算机应用》2014年第04期摘要：退化图像复原的关键在于点扩散函数（PSF）的准确估计，针对散焦模糊图像点扩散函数参数未知的情况，提出一种基于图像边缘模糊频谱特征的参数估计方法。
首先分析基本边缘经模糊退化后的频谱特征，进而构建了自然图像的边缘模型作为参考图像，通过在连续的散焦值范围内计算与待测模糊图像频谱的最大相似性，以获取散焦参数估计值。
实验结果表明，所提方法能够适用于大尺度模糊图像的参数估计问题，且具有较强的抗噪性能。
关键词：散焦模糊；图像边缘；参数估计；抗噪性；频谱相关0 引言在成像系统中，由于照相机或摄像机等成像设备对焦不准而导致的散焦模糊图像，降低了其在刑侦、物证鉴定等领域的应用价值。
基于此，采用数字图像处理技术从模糊图像中提取更多的有价值信息具有重要的现实意义。
目前，国内外的许多专家和学者针对散焦模糊图像的复原已经做了大量的研究[1-5]。
采用先估计点扩散函数（Point Spread Function， PSF），然后选择相应滤波器的复原方法，由于其结构简单、运算高速而被广泛使用。
这类方法的有效性取决于PSF参数的准确估计。
具体到散焦模糊参数的估计，代表性的方法有：利用模糊图像显著的频谱特征，在频域检测零点的方法[6-7]；在限定的参数取值范围内，遍历寻找最优解的方法[8-10]；基于倒谱域极小值点与散焦模糊参数关系的估计方法[11-12]；二次模糊频域相关性方法[13]以及带参考图频谱相关的参数估计法[14]。
零点检测法和倒谱域方法受噪声的影响较大，最优解法的有效性取决于最优判据的选择，相关系数分析方法可以较少的计算量达到较好的自动识别精度，基于参考图像频谱相关的方法区别于以往由模糊图像自身提取信息估计参数，转化为与参考图像的匹配问题，达到了较高的精度。
本文提出一种基于图像边缘模糊频谱特征的散焦模糊参数估计方法，可以适用于大尺度模糊图像的参数估计问题，且抗噪性能进一步得到提升。
盲源分离应用领域
盲源分离应用领域
盲源分离（BSS: Blind Source Separation），又称为盲信号分离，是指在信号的理论模型和源信号无法精确获知的情况下，如何从混迭信号（观测信号）中分离出各源信号的过程。
盲源分离和盲辨识是盲信号处理的两大类型。
盲源分离的目的是求得源信号的最佳估计，盲辨识的目的是求得传输通道的混合矩阵。
应用领域
盲源信号分离是一种功能强大的信号处理方法，在生物医学信号处理，阵列信号处理，语音信号识别，图像处理及移动通信等领域得到了广泛的应用。
盲源分离（BSS：Blind source separation），是信号处理中一个传统而又极具挑战性的问题，BSS指仅从若干观测到的混合信号中恢复出无法直接观测的各个原始信号的过程，这里的“盲”，指源信号不可测，混合系统特性事先未知这两个方面。
在科学研究和工程应用中，很多观测信号都可以看成是多个源信号的混合，所谓鸡尾酒会
问题就是个典型的例子。
其中独立分量分析ICA（Independent component analysis）是一种盲源信号分离方法，它已成为阵列信号处理和数据分析的有力工具，而BSS比ICA适用范围更宽。
目前国内对盲信号分离问题的研究，在理论和应用方面取得了很大的进步，但是还有很多的问题有待进一步研究和解决。
独立成分分析（IndependentComponentAnalysis）
独⽴成分分析（IndependentComponentAnalysis）1. 问题：1、上节提到的PCA是⼀种数据降维的⽅法，但是只对符合⾼斯分布的样本点⽐较有效，那么对于其他分布的样本，有没有主元分解的⽅法呢？2、经典的鸡尾酒宴会问题（cocktail party problem）。
假设在party中有n个⼈，他们可以同时说话，我们也在房间中⼀些⾓落⾥共放置了n个声⾳接收器（Microphone）⽤来记录声⾳。
宴会过后，我们从n个麦克风中得到了⼀组数据{x i x1(i),x2(i),…,xn(i);i=1,…,n}，i表⽰采样的时间顺序，也就是说共得到了m组采样，每⼀组采样都是n维的。
我们的⽬标是单单从这m组采样数据中分辨出每个⼈说话的信号。
将第⼆个问题细化⼀下，有n个信号源s(s1,s2,…s n)T,S∈R n,每⼀维都是⼀个⼈的声⾳信号，每个⼈发出的声⾳信号独⽴。
A 是⼀个未知的混合矩阵（mixing matrix），⽤来组合叠加信号s，那么x=Asx的意义在上⽂解释过，这⾥的x不是⼀个向量，是⼀个矩阵。
其中每个列向量是x(i)，x(i)=As(i)表⽰成图就是这张图来⾃/doc/7be7c1dbce2f0066f53322ed.html /research-interests/research-inte rests-erp-analysis/blind-source-separation-bss-of-erps-using-indepe ndent-component-analysis-ica/x(i)的每个分量都由s(i)的分量线性表⽰。
A和s都是未知的，x是已知的，我们要想办法根据x来推出s。
这个过程也称作为盲信号分离。
令W=A?1，那么s(i)=A?1x(i)=Wx(i)将W表⽰成其中，其实就是将W i 写成⾏向量形式。
那么得到：s j(i )=w j T x (i )2. ICA 的不确定性（ICA ambiguities ）由于w 和s 都不确定，那么在没有先验知识的情况下，⽆法同时确定这两个相关参数。
一种基于类内类间距离的ICA特征选择方法
一种基于类内类间距离的ICA特征选择方法作者：谢勤岚胡晓勤来源：《现代电子技术》2009年第21期摘要:独立分量分析(ICA)可以实现特征提取,但不能直接用于特征选择。
对数据进行ICA 后得到混合矩阵和独立分量,独立分量可以作为特征矢量,混合矩阵可以用于进行特征选择。
首先,使用一种距离度量来计算混合矩阵每一类的类内类间距离比;然后对每一类按该比值由小到大重新排列混合矩阵和独立分量,保留权重矩阵中类间类内距离比大的列,及其对应的特征向量;最后对这些特征向量使用遗传算法选择最优特征组。
两个实验验证了该方法的有效性。
关键词:独立分量分析(ICA);类内距离;类间距离;特征选择;遗传算法中图分类号:TP391.4 文献标识码:A文章编号:1004-373X(2009)21-105-04Feature Selection Method Based on ICA and Distance Ratio of within/between ClassXIE Qinlan1,HU Xiaoqin2(1.College of Electrical and Information Engineering,South-Central University for Nationalities,Wuhan,430074,China;2.Wuhan Traffic School,Wuhan,430074,China)Abstract:ICA can implement feature extraction,but it can′t directly fit f or feature selection.The results after ICA are mixing matrix and independent components,the latter is regarded as feature vectors,while the mixing matrix can be used for feature selection.First,the distance ratio between within-class and between-class of each column of mixing matrix is computed by using a distance measurement.Then,for each class,the mixing matrix and independent components are re aligned by sort ascending according to the distance rate,and the columns of the weight matrix with smaller distance ratio and the corresponding features are reserved.Finally,the best feature set is selected by genetic algorithm from these foregoing feature vectors.Two experiments show that the proposed method is valid.Keywords:independent component analysis;within-class distance;between-class distance;feature selection;genetic algorithm0 引言独立分量分析(Independent Component Analysis,ICA)是近年来由盲信源分解(Blind Signal Separation,BBS) 技术发展来的多维信号处理方法,主要用于揭示和提取多维统计信号中的潜在成分[1-3]。
一种基于盲源分离的DOA估计方法
收稿日期 2007-09-06 收修改稿日期 2008-04-08 Received September 06, 2007; in revised form April 8, 2008 1. 海军大连舰艇学院信号与信息技术研究中心, 大连 116018 1. Research Center of Signal and Information, Dalian
矩阵阵R¯ˆX进行= 奇(1/异L值) 分Lτ解=1R[¯ˆXτ
(X¯ )]. 对得到的平均时间延迟相关 = U¯ RS V¯ H + σ2I, 其中 (·)H 表示
共轭转置, U¯ 和 V¯ 分别称为左右奇异矩阵. 则此时估计的阵
列流形 Aˆ = Q#U¯ = [aˆ1, aˆ2, · · · , aˆN ], 其中 (·)# 表示求伪逆,
离, 完成对阵列流形和源信号的估计, 然后根据 ESPRIT 方 法原理, 利用子阵间的相位延迟完成目标方位估计. 仿真实验 和海上实测数据结果表明, 所提方法能准确估计出目标的方 位和相应的源信号. 同时, 在阵元数、信噪比、快拍数的要求 和方位分辨率上表现出了比多重信号分类 (Multiple signal classification, MUSIC) 方法更优的性能.
4.1 仿真实验
仿真实验考虑均匀线列阵位于两窄带目标源的远场, 目 标源 1 的方位为 15 度, 目标源 2 的方位为 16 度. 用式 (9) 所 定义的圴方根误差 (Root mean square error, RMSE) 来检 验所提方法的有效性[3], 并与经典的高分辨方位估计 MUSIC 方法进行对比.
复数算法或修改的实数算法, 完成对阵列接收信号的分离,
如果能实现对阵列流形 A 的估计, 则完全可实现目标方位 θi 的估计.
ITERATIVE TECHNIQUES FOR BLIND SOURCE SEPARATION
EUSIPCO ’92ITERATIVE TECHNIQUES FOR BLIND SOURCE SEPARATIONUSING ONLY FOURTH-ORDER CUMULANTSJean-François CardosoTélécom Paris.Dept Signal,46rue Barrault,75634Paris CEDEX13,FRANCE.Email:cardoso@sig.enst.frAbstract."Blind source separation"is an array processing problem without a priori information(no array manifold).This model can be identified resorting to4th-order cumulants only via the concept of4th-order signal subspace(FOSS)which is defined as a matrix space.This idea leads to a"Blind MUSIC"approach where identification is achieved by looking for the(approximate)intersections between the FOSS and the manifold of1D projection matrices.Pratical implementations of these ideas are discussed and illustrated with computer simulations.1.INTRODUCTIONThis paper adresses the problem of blind source separation or independent component analysis in the complex case where it can be seen as a narrow-band array processing problem:the output,denoted x(t),of an array of m sensors listening at n discrete sources takes the form: (1)x(t)=p=1,nΣs p(t)a p+n(t)where s p(t)denotes the complex signal emmited by the p-th source;where a p is afixed(deterministic)vector called the p-th source signature;and where n(t)is an independent additive noise assumed to be normally distributed with arbitrary covariance.In"standard"array processing,the same data model is used but the source signatures depend on very few location parameters and this dependence is assumed to be known via the array manifold.In contrast,we adress here the blind problem where no a priori information is available about source signatures.Hence blind qualifies any processing based on the sole obervations."Blindness"is compensated by exploiting the hypothesized assumption of source independence.In the following it is assumed that:•source signals are statistically independent,•source signals have non vanishing kurtosis,•source signatures are linearly independent.Blind source separation is understood as estimation of the source signals s p(t),while blind identification refers to the estimation of the source signatures a p.In the following, we focus on identfication since separation can be based on signature estimates.The blind problem is interesting because its solution allows to process narrow band array data without explicit knowledge about array geometry and without assumptions about wavefront shapes.Various solutions relying on the use of both2nd-order and4th-order cumulant statistics of the array output have already been reported[1-4].These approaches make use of2nd-order information to whiten the data,and4th-order information is then used to process the resulting orthogonalized problem.The additive noise is assumed to be normally distributed:it has no effect on the(exact)4th-order cumulants but2nd-order noise cumulants do not vanish so that the spatial structure of the noise covariance has to be known,modelled or estimated in order to achieve consistent2nd-order prewhitening. These limitations can be overcome by giving up the idea of 2nd-order whitening and resorting to4th-order cumulants only[5-6].It is the purpose of this communication to show how the concept of fourth-order signal subspace yields simple implementations of4th-order only blind identification.2.BLIND IDENTIFICATION2.1On identifiability.The blind context does not lead to full identifiability of the model(1)because any complex factor can be exchanged between s p and a p without modifying the observation.Hence if no a priori information is available,each signature can be identified only up to a scale factor.We take advantage of this to assume,without any loss of generality,that each signature a p has unit norm.With this constraint,an unidentifiable phase term is still present in a p.For each source p,we denote asΠp the orthogonal projector onto the1D space where the p th component lives.It is the space spanned by the signature a p and the projector onto it is(2)Πp=∆a p a p∗This hermitian matrix is unaffected by any phase term in a p and conversely determines a p up to a phase term.It follows that the projectorsΠp are the algebraic quantities that can be,at best,identified in the blind context.It iseasily seen that knowing these projectors is sufficient to perform blind separation of the source signals s p(t) because they allow to construct for each source the linear filter that zeroes all the components but the one specified. We then define blind identification as the problem of estimating the projectorsΠp from sample statistics only. 2.2Blind MUSIC.The approach to blind identification presented in this contribution can be seen as a blind4th-order version of the celebrated MUSIC algorithm.The MUSIC technique is based on the concept of signal subspace which is the vector space spanned by the steering vectors.It can be summarized as i)Estimate the signal subspace using the covariance.ii)Search for the steering vectors which are the closest to the signal subspace.The search is,of course,across the array manifold:this is how MUSIC exploits the a priori information contained in the parameterization of the steering vectors.The fourth-order signal subspace(FOSS)is defined as the real span of the projectorsΠp i.e.as the linear matrix space made of all possible linear combinations with real coefficients of the projectorsΠp:(3)FOSS={M|M=p=1,nΣγp a p a p∗,γp∈R}Let us now consider the following idea for blind identification i.e.estimation of theΠp:i)Estimate the FOSS using4th-order cumulants.ii)Search for the orthogonal1D projectors which are the closest to the FOSS.The closest projectors to the FOSS are taken as estimates of the source projectorsΠp.Such an idea could be termed"blind MUSIC"because,in spite of its strong analogy with the classical MUSIC,no signature parameterization is assumed here:the search is across the so called rank one manifold(ROM)which is defined as the set of all rank-one unit-norm hermitian matrices i.e.across all the1D orthogonal projectors.The reason why Blind MUSIC works is that a matrix space with structure as in eq.(3)is shown,under mild conditions,to contain no other1D projectors than the ones used in its construction,i.e.theΠp.This is obviously true as soon as the signatures a p are linearly independent since in that case,a matrix M as in eq.(3)has a rank equal to the number of non-zero coefficientsγp.If M is a1D projector, it has rank one,hence all the coefficientsγp but one are zero and M is then necessarily one of theΠp.We now have to discuss i)FOSS estimation from4th-order cumulants ii)practical implementations of the blind MUSIC search.3.FOURTH-ORDER SIGNAL SUBSPACE3.1Quadricovariance.Wefind convenient to make temporary use of indexed notations to express the cumulants of the vector process x.Let us denote by x i the i-th coordinate of vector x and by x i the i-th coordinate of its dual x∗.Of course,only orthonormal basis are used: x i is just the complex conjugate of x i.The covariance classically is the matrix R whose(i,j)-coordinate denoted r i j is the2nd-order cumulant of x i and x j:(4)r i j=∆Cum(x i,x j)1≤i,j≤m Similarly,we define the quadricovariance of x as the set of m4complex scalars,q il jk,1≤i,j,k,l≤m:(5)q il jk=∆Cum(x i,x j,x k,x l)Our approach to process4th-order information is to consider the quadricovariance as a matrix mapping denoted Q,which to any matrix M with coordinates m i j associates the matrix N=Q(M)with coordinates n i j according to:(6)n i j=1≤k,l≤mΣq il jk m k lThe quadricovariance has the two following properties:i) it maps any hermitian matrix to another hermitian matrix. ii)it is itself an hermitian operator in the(usual)sense that for any matrices M and N we have<N|Q(M)>∗= <M|Q(N)>with the Euclidian scalar product <M|N>=∆Tr(NM H).These are trivial consequences of cumulant symmetries.It follows[3]that the quadricovariance admits m2real eigenvalues,denoted µi,i=1,m2and m2corresponding orthonormal hermitian eigen-matrices,denoted E i,i=1,m2,verifying:∀i=1,m2Q(E i)=µi E i withE i=E i H,µi∈R,Tr(E i E j)=δ(i,j)As a simple consequence[3]of cumulant additivity and multilinearity,the quadricovariance of a linear mixture(1) of independent components takes the special form:(7)Q(M)=p=1,nΣk p a p∗Ma a p a p a p∗with no contribution from the additive noise(since it has been assumed Gaussian and independent of the signals) and where the kurtosis of the p-th source is denoted by k p: (8)k p=∆Cum(s p,s p∗,s p∗,s p)Equation(7)evidences that the image space of the quadricovariance Q is spanned by the projectors Πp=a p a p∗(hence the name"FOSS").It has exactly rank n if no kurtosis k p is zero and if the projectorsΠp are linearly independent.This last condition is fulfilled whenever the signatures a p are themselves independent.It follows that quadricovariance eigen-decomposition shows only n non-zero eigenvalues.Let us assume that they are numbered in such a way that the corresponding n eigen-matrices are(E i|i=1,n).These eigen-matrices form an hermitian orthonormal basis of the FOSS.3.2FOSS estimation.When a strongly consistent estimate of the signal covariance is used,the2nd-order signal subspace estimate obtained via eigen-decomposition also is strongly consistent.The same can be shown to hold for the FOSS estimates obtained from an eigen-decomposition of the sample quadricovariance into eigen-matrices.This should be the preferred FOSS estimation method for small arrays but eigen-decomposition of the quadricovariance may be too expensive with large arrays.Note however that only a small number of eigen-matrices need to be estimated(n and not m2)and this fact can lead to large computational savings(see[7]).Even in that case, the whole set of4th-order cumulants is needed,and quadricovariance estimation cost may be prohibitive. Fortunately,the FOSS can be estimated in a simpler manner.We demonstrate this in the(rather common)case where signals are circurlarly distributed.Cumulant expression in terms of the moments then reduces to: (9)q il jk=E{x i x j x k x l}−r i j r l k−r l j r i kand it is readily checked that the quadricovariance image of any matrix M accordingly reduces to:(10)Q(M)=E{(x∗Mx x)x x∗}−RMR−R Tr(MR) This expression admits an obvious sample counterpart showing that Q(M)can be estimated at a cost similar to the covariance.This suggests to choose a priori a set of n hermitian matrices M i and to estimate Q(M i)according to (10).The result will be a set of n almost surely independent matrices of the FOSS.They can then be orthonormalized(by a Gram-Schmidt procedure for instance)into an orthonormal hermitian basis of the FOSS. Such a procedure obviously yields FOSS estimates with higher variance than those obtained by eigen-decomposition of the whole set of4th-order cumulants. Since it is not possible to ensure in advance that the Q(M i) actually are independent,a safer solution would be to use a number of M i larger than n.4.BLIND MUSIC IMPLEMENTATIONFrom now on,we assume that a FOSS estimate is available in the form of a set of n hermitian orthonormal matrices:(M i|i=1,n)forming a basis of the estimated FOSS.The following search implementations do not depend on the particular FOSS estimation technique.Blind MUSIC can be implemented as searching through the FOSS the closest ROM matrix(see the PQN technique below)or,alternatively,as searching through the ROM the closest FOSS matrix(seeΠV3)In both cases,the suggested techniques do not implement the search via a gradient(or similar)approach but expresses the Blind MUSIC estimates asfixed points of an appropriate mapping.In our simulations,we have found that thesefixed points were the only stable points:the blind MUSIC search can then be implemented as the iteration of these mappings with arbitrary starting points.4.1ΠV3Search.The natural approach to blind MUSIC is to maximize the norm of the projection onto the FOSS of a matrix A under the constraint that it is a1D projector. Using an orthonormal hermitian basis,the squared norm of this projection,denoted d,is the sum of the squared projections onto each basis matrix.(11)d=i=1,nΣTr2(A M i)Blind MUSIC estimates are obtained as the maximizers of d under the constraint that A=v v∗with v∗v=1.Since Tr(A M i)=v∗M i v,the variation with respect to v of a Lagrange function L=1/2d−λv∗v associated to this constrained optimization problem is:δL=i=1,nΣ(v∗M i v)(v∗M iδv+δv∗M i v)−λ(v∗δv+v∗δv) Defining the cubic vector mapping v→φ(v)as: (13)φ(v)=∆Σi=1,n(v∗M i v)M i vthe Lagrange function variation is rewritten in:(14)δL=(φ−λv)∗δv+δv∗(φ−λv)which is zero for anyδv iffφ(v)=λv.This is equivalent to v being afixed point of the mapping v→Φ(v)where: (15)Φ(v)=∆φ(v)/|φ(v)|TheΠV3search(where V3is a reminder for the cubic dependence on the iterated vector)starts with a random vector and then iteratively computes its image throughΦ.4.2PQN Search.An alternate approach is to search for matrices of the FOSS that are as close as possible to the ROM.The basic idea is to start with an arbitrary matrix of the FOSS and to repeatedly project it onto the ROM and back onto the FOSS.Projection onto the ROM is equivalent to truncating the matrix to itsfirst principal eigen-component,which requires an eigen-decomposition at each step.On the other hand,repeatedly squaring a matrix has the effect of enhancing the dominant eigenvalue.In our experiments,we have found that the projection onto the ROM,being included in the iteration loop,could be replaced by a simple matrix squaring followed by renormalization The PQN algorithm is just cycling through the three steps of projection,quadration, and normalization,hence the acronym"PQN".After convergence,the dominating eigen-vector is extracted, providing an estimate of one of the source signatures.Quadration and projection can be efficiently implemented in a single step by representing the iterated matrix,say A,by its(real)coordinates a i,i=1,n in the FOSS basis:A=Σa i M i.The squared matrix then is A2=Σa i a j M i M j.Since an orthonormal basis is used, the projection of A2onto the FOSS has coordinates a k′given by:(16)a k′=Σi,j t ijk a i a j with t ijk=∆Tr(M i M j M k) Hence,by pre-computing the table t ijk,each quadration-projection is computed in the single step(16),involving only n3real multiplications.4.3Simulation results The following simulation results are for a uniform linear half-wavelength array of4sensors, two independent PSK modulated sources of unit variance located respectively at0and20degrees(i.e.under the same lobe).The signal is corrupted by additive Gaussian white noise with covarianceσI.We performed20Monte-Carlo runs using the PQN algorithm for data lengths of50, 100,200,500,1000and for noise levelsσ=-10,0,10,20 dB.For each run,we plot a performance indexρdefinedas (17)ρ=n1p =1,nΣ1−Tr (Πˆp Πp)where each Πˆpis the estimated p -th projector.This index also is the squared sine of the angle between each signature and its estimate (averaged on the sources).Perf.index10E-110E-210E-310E-4501002005001000-10dB SNR•••••••••••••••••••••••••••••••Perf.index10E-110E-210E-310E-45010020050010000dB SNR•••••••••••••••••••••••••••••••••••••••••••••••••Perf.index10E-110E-210E-310E-450100200500100010dB SNR•••••••••••••••••••••••••••••••••••••••••••••••••••••••••••For negative SNR,the data lengths used here do notallow any correct estimation :the performance index is close to 0.5,indicating "random estimation"!But as the SNR gets positive meaningful estimates are obtained with relatively short data lengths.Also note that there is no significant improvement when the SNR goes from 10dB to 20dB.This is a general feature of 4th-order-only blind techniques :when the noise level is low enough,the performance is dominated by the sample size.This could be contrasted with the standard parametric MUSIC (2nd-order or 4th-order)where,at low noise levels,the varianceSample sizePerf.index10E-110E-210E-310E-450100200500100020dB SNR •••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••of the estimates is proportional to the noise power.Another remark is that these simulations are for equipowered sources :the performance would degrade for a source when its power gets weaker.This effect shows at any order but is naturally more severe at fourth-order than at 2nd-order.Just as in the MUSIC case,asymptotic performance can be obtained in closed form but room is definitely lacking for exposition of the results.They will be presented at the conference.CONCLUSION.The notion of fourth-order signal subspace (FOSS)has been introduced.This matrix space,function of the 4th-order cumulants,is the natural 4th-order counterpart of the classical 2nd-order signal subspace.Exploited in a "blind MUSIC"fashion,it allows for blind identification without resorting to 2nd-order information.This can be done with one of the low cost fixed point techniques presented here.REFERENCES[1]on,"Independent Component Analysis",Proc.Int.Workshop on Higher-Order Stat.,Chamrousse,France,Jul.91,pp.111-120.[2]M.Gaeta,coume,"Source Separation Without A Priori Knowledge :the Maximum Likelihood Solution",Proc.EUSIPCO,Barcelona,Spain,Sept.90,pp.621-624.[3]J.F.Cardoso,"Eigen-structure of the fourth-order cumulant tensor with application to the blind source separation problem",Proc.ICASSP’90,pp.2655-2658,Albuquerque,1990.[4]V.C.Soon,L.Tong,Y.F.Huang,R.Liu,"An extended fourth order blind identification algorithm in spatially correlated noise",Proc.ICASSP’90,pp.1365-1368,Albuquerque,1990.[5]J.F.Cardoso,"Super-Symmetric Decomposition of the Fourth-Order Cumulant Tensor.Blind Identification of More Sources than Sensors",Proc.ICASSP’91,pp.3109-3112,Toronto,1991.[6]G.Giannakis,S.Shamsunder,"Modelling of non Gaussian array data using cumulants :DOA estimation with less sensors than sources",Proc.of conf.on Info.Sci.and Syst.,Baltimore,MD,1991.[7]J.F.Cardoso,on,"Tensor-based Independent Component Analysis",Proc.EUSIPCO,Barcelona,Spain,Sept.90,pp.673-676.。
配管词汇大全中英文对照
"s" bend s形弯管"U" bend U形弯管“A-A”剖视Section “A-A”“S” bend S形弯管“U” bend U形弯管“X”视图|View “X”20 alloy|20合金20合金|20 alloy3-piece mitre bend|三节斜接弯管3-way ball valve|三通球阀45 degree lateral|45度斜三通45 degree lateral (reducing on one run and branch)|45度斜三通（一个直通口及支管为异径）45 degree lateral (reducing on one run)|45度斜三通（一个直通口为异径）45 degree lateral(reducing on branch)|45度斜三通（支管为异径）45° lateral|45°斜三通45°lateral(reducing on branch)|45°斜三通（支管为异径）45°lateral(reducing on one run and branch)|45°斜三通（一个直通口及支管为异径）45°lateral(reducing on one run)|45°斜三通（一个直通口为异径）45°斜三通（一个直通口及支管为异径）|45°lateral(reducing on one run and branch) 45°斜三通（一个直通口为异径）|45°lateral(reducing on one run)45°斜三通（支管为异径）|45°lateral(reducing on branch)45度斜三通|45° lateral45度斜三通（一个直通口及支管为异径）|45degree lateral (reducing on one run and branch)45度斜三通（一个直通口为异径）|45degree lateral (reducing on one run)45度斜三通（支管为异径）|45degree lateral(reducing on branch)8字盲板|Spectacle blind ; figure 8 blind90°弯管|quarter bend90度弯管|Quarter bendAbove ground piping|地上管道Absolute elevation|绝对标高Absorber|吸收塔access way|走道；过道Accessory|附件；附属设备Accuracy of take-off|统计材料准确度Acetal plastic|缩醛塑料Acid-proof paint|耐酸漆Acoustic vibration|音响振动Acrylic resin|丙烯酸树脂Actuator|执行机构Adhesive|胶粘剂Adjustable cleat|可调夹板Adjustable support|可调支架Adsorber|吸附器Advanced certified final|先期确认After cooler|后冷却器Ageing treatment|时效处理Agitator|搅拌器Air|空气Air cooler|空冷器Air separation facility|空分装置Air tightness test|气密试验Air vent valve|放气阀Alignment|对中心；找正Alignment tolerance|错边量Alkali-proof paint|耐碱漆Alkyd enamel|醇酸瓷漆Allowable stress|许用应力Allowable stress range|许用应力范围Allowance|允差Alloy steel|合金钢alloy steel pipe|合金钢管Alloy steel pipe|合金钢管Alternating current|交流Alternating stress|交变应力Altitude|高度；海拔Aluminium sheet|铝板Aluminizing|渗铝Aluminosilicate fiber|硅酸铝纤维Aluminum|铝Aluminum bronze|铝青铜Aluminum magnesium|铝镁合金Ambient temperature|环境温度American standard taper pipe thread (NPT)|美国标准锥管螺纹American wire gage|美国线规Ammonia gas|氨气Ampere|安（培）Amplitude|振幅；波幅Analytical engineering phase|分析设计阶段Analyzer room|分析室Anchor|固定架Anchor bolt|地脚螺栓anchor point|固定点Angle steel|角钢Angle valve|角阀Angular rotation|角位移Annealing|退火Anti-corrosive paint|防腐漆Antirust paint|防锈漆Anti-sweat|防结露Apparatus|仪器；设备Appendant displacement|附加位移Appendix|附件Approval|审定Approved for construction|批准用于施工Approved for design|批准用于设计Approved for planning|批准用于规划布置Approximate|大约Arc welding|电弧焊Arcylonitrile-butadiene-styrene|丙烯腈-丁二烯-苯乙烯Area|面积；区域Area limit|区界Argon-arc welding|氩弧焊Arithmetical average roughness height (AARH)|算术平均粗糙高度As built drawing|竣工图Asbestos board|石棉板Asbestos cloth|石棉布asbestos cord|石棉绳Asbestos emergency seal|石棉安全密封Asbestos fabric|石棉织品Asbestos rope|石棉绳Asbestos rope with inconel wire|带铬镍合金丝的石棉绳Asphalt|沥青Asphalt felt|油毛毡；沥青毡Assembly|组装；装配Atmosphere|大气压Atmospheric pollution|大气污染Attachment of support|管架零部件attenuation constant|衰减系数Austenitic stainless steel|奥氏体不锈钢austenitic stainless steel pipe|奥氏体不锈钢管Austenitic stainless steel pipe|奥氏体不锈钢管Automatic analysis|自动分析Automatic submerged arc welding|自动埋弧焊Auxiliary boiler|辅助锅炉Axial movement type expansion joint|轴向位移型膨胀节Axial pump|轴流泵Axial stress|轴向应力Back pressure regulating valve|背压调节阀Back run welding|封底焊Back seal|倒密封Back to back|背至背Backing weld|垫板焊Balanced pressure thermostatic trap|压力平衡式恒温疏水阀Ball check valve|落球式止回阀Ball type expansion joint flexible ball joint|球形补偿器Ball valve|球阀Bar|巴Bare line|裸管Barometric leg|大气腿base elbow|带支座弯头Base elbow|带支座弯头Base metal|母材；基层金属Base plate|底板base tee|带支座三通Base tee|带支座的三通Basic design|基础设计Basic engineering design data|设计基础数据Battery limit|项目区界Battery limit condition|界区条件Battery room|蓄电池室Bead welding|珠焊Beam|梁Bearing|轴承Bell and spigot joint|承插连接（接头）Bell end|承口Bellow expansion joint|波纹膨胀节Bellow sealed valve|波纹管密封阀Belt conveyor|皮带输送机Bend|弯管Bending moment|弯曲力矩Bending strength|抗弯强度Bending stress|弯曲应力Bending test|弯曲试验Bevel for combined thickness|内外侧厚度切斜角Bevel for inside thickness|内侧厚度切斜角Bevel for outside thickness|外侧厚度切斜角Bevel gear operated|伞齿轮传动Beveled end|坡口端bib|（水）龙头Bibb|（水）龙头Bid|投标Bill of material|材料表Bimetallic expansion steam trap|双金属膨胀式蒸汽疏水阀bin|料仓Bituminous paint|沥青漆Blank|插板Blanket|棉毡Bleed valve|抽出液阀（小阀）blind|法兰盖blind flange|法兰盖Blind flange|法兰盖Blister|砂眼Block form|分组（类）形式（如管道等级表）Block valve|切断阀Blow down|排污Blow down tank|排污罐blow hole|气孔Blowdown valve|排污阀Blower|鼓风机Blow-off|吹出Blowout proof stem|防脱出阀杆Blue|蓝色的Body|阀体Body stem seal|本体阀杆密封Boiler|锅炉Boiler feed water|锅炉给水Boiler feed water pump|锅炉给水泵Bolt|螺栓Bolt circle|螺栓圆Bolted bonnet|螺栓连接的阀盖Bolted cap|螺栓连接的阀帽Bonnet|阀盖Bonnet bush|阀盖衬套Bonnet gasket|阀盖垫片Booster|升压器Borosilicate glass|硼硅玻璃boss|支管台，插入式支管台Boss|支管台，插入式支管台Both end plain|两端平Both end thread|两端带螺纹bottom coat|底漆Bottom of pipe|管底Bottom of trench|沟底Bracing|斜撑；支撑Brackish water|苦咸水（碱性水）Branch connection|支管连接Branch pipe welded directly to the run pipe|焊接支管Brass|黄铜Braze welding|硬钎焊Breather valve|呼吸阀Bridge crane|桥式起重机Brimingham wire gage|伯明翰线规Brine|盐水Brinell hardness|布氏硬度Brittleness|脆性Bronze|青铜Brown|棕色的；褐色的Build up welding|堆焊Building|建筑物Bulk material|散装材料Burn through|焊穿Burner|烧咀bushing|内外螺纹缩接（俗称补芯）Bushing|内外螺纹缩接（俗称补芯）Butt welded|对焊的Butt welded end|对焊端Butt welded joint|对焊连接（接头）Butt welding|对焊Butterfly check valve|蝶式止回阀Butterfly valve|蝶阀By buyer|由买方提供By pass|旁路By seller|由卖方供货By vendor|由制造厂供货By-pass valve|旁路阀Cable|钢索，电缆cable rack|电缆槽（架）Cable tray|电缆槽（架）Cable trench|电缆沟Calcium silicate|硅酸钙Calculation sheet|计算书Candtilever support|悬臂架Canted disc butterfly valve|斜阀盘蝶阀Cap|管帽cap[c]|管帽Capacitance|电容Car seal close|未经允许不得开启Car seal open|未经允许不得关闭Carbon steel|碳素钢carbon steel pipe|碳钢管Carbon steel pipe|碳钢管Carbonization|渗碳Cast iron|铸铁cast iron pipe|铸铁管Cast iron pipe|铸铁管Cast steel|铸钢Cast valve|铸造阀Casting|铸件Cat walk|小过道cat way|小过道Catch basin|集水池Cathodic protection|阴极保护Caulking material|填颖材料Caustic soda|烧碱Cavitation erosion|气蚀cellular concrete|泡沫混凝土cellular glass|泡沫玻璃Cellular polystyrene|泡沫聚苯乙烯Cellulose acetate butyrate|醋酸丁酸纤维素Celsius|摄氏Center line|中心线Center line of discharge|出口中心线Center line of suction|入口中心线Center to center|中心至中心；中至中Center to end|中心至端面Center to face|中心至面Centrifugal compressor|离心式压缩机Centrifugal filter|离心过滤机Centrifugal pump|离心泵Centrifugal separator|离心分离机Centrifuger|离心机Ceramic|陶瓷Certified final (CF)|最终确认Chain block|手动葫芦；倒链Chain operated|链条操纵的Chain wheel|链轮Channel|槽钢Check|校核Check list|审核提纲Check valve|止回阀Checkered plate|网纹钢板Chemical analysis|化学分析Chemical cleaning|化学清洗Chemical sewage|化学污水Chiller|深冷器Chlorhydric acid|盐酸Chlorinated polyether|氯化聚醚Chlorinated polyvinyl chloride|氯化聚氯乙烯Chrome-molybdenum steel|铬钼钢chrome-nickel steel|铬镍钢chrome-plated|镀铬的Chromium steel|铬钢Chromium-nickel steel|铬镍钢Chromium-plated|镀铬的Chromized steel|渗铬钢，镀铬纲Chromizing|渗铬circle bend|环形弯管Circle bend|环形弯管Circulating compressor|循环压缩机Circulating water|循环水Circulation|循环Circumferential band|环箍Circumferential stress|圆周应力Clad|金属保护层clad pipe|复合管Clad pipe|复合管Clad steel|复合钢cladding metal|金属保护层Clamp|管卡Clarifier|沉淀池Class|压力级；等级；类别Class designation|管道等级号Clean out|清扫口Clearance|间隙Cleat|夹板，导向板Clevis|U形夹（卡）Client change notice|用户变更通知Client customer|顾客Clip|锚固件；生根件;预焊件Clip on equipment|预焊件（设备上）Clock wise|顺时针方向Closet|盥洗室；厕所Coarse|粗制的Coarse thread|粗牙螺纹Coating|涂层；覆盖层Cock|旋塞Code number|代码Coefficient of stress concentration|应力集中系数Cold flow|冷流Cold insulation|保冷Cold load|冷态荷载cold quenching|水冷淬火Cold rolling|冷轧Cold shortness|冷脆Cold spring|冷拉Cold working|冷加工Cold-drawing seamless pipe|冷拔无缝钢管cold-drawing seamless pipe|冷拔无缝钢管Colour|颜色Column|柱column|塔Combination U and V groove|U-V组合坡口Commissioning|试车Compact type|紧凑型（小型）Companion flange|配对法兰Compressed asbestos gasket|压缩石棉垫片Compression stress|压应力Compressive strength|抗压强度Compressor|压缩泵Compressor house|压缩机房Computer aided design|计算机辅助设计Concentration|浓度concentric reducer|同心异径管Concentric reducer|同心异径管Condenser|冷凝器Conduit|导管Conduit tube|导线管Confirm plot plan|确认版设备布置图（“D”版）Connecting rod|连接杆Consolidated piping material summary sheet|综合管道材料表Constant hanger|恒力吊架Construction|建设Construction plot plan|施工版设备布置图（“G”版）Contact corrosion|接触腐蚀Contaminated rain water|污染雨水Contamination|污染Continue on drawing|接续图Contraction|承包商Control room|控制室Control valve|控制阀Convection section|对流段Conventional heat treatment|普通热处理Converter|转化器；变换器Conveyor|输送机Cooling|冷却水Cooling tower|冷却塔Cooling water return|循环冷却水回水Cooling water supply|循环冷却水给水Coordinate|坐标Copolymer|共聚物Copper|铜；紫铜Cork wood|软木Corrosion|腐蚀Corrosion allowance|腐蚀裕量Corrosion inhibitor|防腐剂Corrosion resistance|耐蚀性Corrosion test|腐蚀试验corrugated bend|折皱弯管Corrugated bend|折皱弯管Corrugated metal double jacketed asbestos filled gasket|双夹套波纹金属包石棉垫片Corrugated metal gasket|波纹金属垫片Corrugated metal gasket with asbestos inserted|波纹金属包嵌石棉垫片Cotter pin|开口销Counter clock wise|逆时针方向Counter weight hanger|重锤式吊架Countersunk head screw|沉头螺栓Couple of force|力偶coupling|管接头Coupling|管接头Cradle|托架Crane|起重机Crate|板条箱crater|弧坑Creep limit|蠕变极限Creep rupture strength|蠕变断裂强度Crevice corrosion|缝隙腐蚀Critical piping|重要管道Critical point|临界点Critical pressure|临界压力Critical temperature|临界温度Cross|四通cross-over bend|跨越弯管（︹形）Cross-over bend|跨越弯管Cryogenic service valve|低温用阀Crystallizer|结晶器Cubic meter|立方米Current|电流customer|顾客Cut to suit|切割使适合……Cyclone|旋风分离器Cylinder|气缸Cylinder operated|气缸（或液压缸）操纵的Damped vibration|阻尼振动Damper|风门、挡板Damping device|减振装置Dark|深色的Dash pot|缓冲筒（器）Data base|数据库Data sheet|数据表Davit|吊柱Dead load|静荷载Dead-soft annealing|极软退火Deaerator|脱氧器decay coefficient|衰减系数Decay factor|衰减系数Decibel|分贝Deep well pump|深井泵Defects of welding|焊接缺陷Deflection|挠度；弯度deformation|应变；变形Degasifier|脱气塔Degree|度Delivery order|交货单Demineralized water|脱盐水Demineralizer|脱盐装置Density|密度Department|部门Deposited seat|堆焊（阀）座Depth|深度Description|说明Design document|设计文件Design manager|设计经理Design note|设计注释Design pressure|设计压力Design response spectrum|设计响应谱Design seismic coefficient|设计震度Design specification summary sheet (DSSS)|设计规定汇总表Design temperature|设计温度Designing plot plan|设计版设备布置图（“F”版）Desulphurization reactor|脱硫反应器Desuperheater|减温器Detail|祥图Detail design|详细设计Detail design issue|详细设计版Diameter|直径diamond penetrator hardness|维氏硬度Diaphragm operated control valve|膜式控制阀Diaphragm valve|隔膜阀diatomaceous earth|硅藻土Differential pressure regulating valve|差压调节阀Dimension|尺寸Direct current|直流Direct-fired heater|回热炉Direction|方向Directional stop|定向限位架Disc|阀盘Disc seat|阀盘密封圈discharge|出口Discharge valve|排出阀Discipline|专业；学科Displacement|位移Displacement stress|位移应力Displacement thermal expansion stress range|位移热胀应力范围Distance|距离Distillation tower|蒸馏塔Diverting valve|换向阀Documentation|资料;文件Door|门Double bellow|双波Double bevel groove|K形坡口double branch elbow|双支管弯头Double branch elbow|双支管弯头Double disc parallel seat|平行双闸板Double extra heavy|双倍加厚的；双倍加强的double extra strong|双倍加厚的；双倍加强的Double jacketed gasket|双夹套垫片double offset expansion "U" bend|双偏置U膨胀弯管Double offset expansion “U” bend|双偏置U膨胀弯管Double U groove|双面U形坡口Double V groove|X形坡口Double-acting limit stop|往复定值限位架Dowel pin|定位销Down|下Drain|排液Drain funnel|排液漏斗Drain valve|排液阀drawing|图Drawing|拔制Drawing number|图号Drawn|制图Drill|钻孔Drip leg|集液包Drip ring|排液环Drip valve|集液排放阀Drum|罐Dry gas-holder|干式气柜Dryer|干燥器Dual plate wafer type check valve|双板对夹式止回阀Ductility|延性Dye penetrant inspection|着色渗透检验Dynamic analysis|动态分析Dynamic load|动力荷载ear|支耳；吊耳Earth lug|接地板earthing|接地Earthquake|地震earthquake load|地震荷载East|东eccentric butterfly valve|偏心阀板蝶阀Eccentric reducer|偏心异径管eccentric reducer|偏心异径管Economizer|省煤器Eddy current test|涡流探伤Ejector|喷射器Elastic limit|弹性极限Elastomer with asbestos fabric insertion|夹石棉织物的橡胶Elastomer with asbestos fabric insertion and with wire reinforcement|夹石棉织物及金属丝加强的橡胶Elastomer with cottan fabric insertion|夹棉织物的橡胶elbolet|弯头支管台Elbolet|弯头支管台elbow|弯头Elbow|弯头Electric fusion welding|电熔焊Electric heater|电加热器Electric motor operator|电动操纵器Electric resistance welding|电阻焊Electrical panel|电气盘Electrical tracing|电伴热Electrically operated valve|电动阀Electric-arc-welded steel-plate pipe|电弧焊钢板卷管electric-fusion(arc)-welded steel-plate pipe|电熔（弧）焊钢板卷管Electric-fusion-welded steel-plate pipe|电熔焊钢板卷管Electric-motor operated valve|电动阀electric-resistance welded steel pipe|电阻焊钢管Electric-resistance welded steel pipe|电阻焊钢管Electro corrosion|电化腐蚀Electroslag welding|电渣焊Elevation|标高；立面Ellipsoidal head|椭圆形封头Embedded part|预埋件Emergency valve|事故切断阀Enamel|瓷漆End connection|端部连接Endurance limit|持久极限Engineering drawing|工程图Engineering manual|工程手册；设计手册Engineering specification|工程规定Epoxy|环氧树脂epoxy resin|环氧树脂Epoxy resin paint|环氧树脂漆equipment and piping arrangement|设备布置及管道布置Equipment item number|设备位号Equipment list|设备表Equipment name|设备名称Equivalent|相当的；当量的Erection|安装Erection opening|吊装孔Error|误差Estimate|估算Estimated price|估价Ethylene perchloride paint|过氧树脂漆;过氯乙烯漆Ethylene propylene diene monomer|乙烯丙烯二烯单体Ethylene propylene rubber|乙丙橡胶Evacuation|插空；排空Evaluation|评标Evaporator|蒸发器Ex pier|码头交货Ex wharf|码头交货excess load|超载Excessive spatter|严重飞溅Excitation|激振；激发Exhaust|排出口Existing steel structure|已有钢结构Expansion bolt|膨胀螺栓Expansion joint|膨胀节Expediting|催货Explosion door|防爆门Explosive welding|爆炸焊External force|外力External pressure stress|外压应力Externally applied load|外载externally imposed displacement|附加位移Extra fine thread|特细牙螺纹Extra heavy|加厚的；加强的extra strong|加厚的；加强的Extractor|萃取器Extruder|挤压机Extruding|挤压Extrusion|伸出长度（指预埋螺栓）Eye bolt|环头螺栓Eye rod|带环头拉杆Eye washer and shower|洗眼器及淋浴器Eye washer station|洗眼站Fabricated pipe bend|预制弯管fabricated pipe bend|预制弯管Face to face|面至面Facilities|设施Facing finish|法兰面加工Fahrenheit|华氏Fan|风机Fatigue limit|疲劳极限Fatigue test|疲劳试验faucet|（水）龙头Feed|进料Feed tank|加料槽Feed water heater|给水加热器Female face|凹面Ferritic alloy steel pipe|铁合金钢管Ferrous metal|黑色金属fiber cloth|玻璃布Fiber reinforced thermoplastics|纤维增强热塑性塑料Field weld|现场焊Figure|图figure 8 blind|8字盲板File|文件Fillet welding|角焊Filter|过滤器Fin|翅片式导向板fine|精制的Fine thread|细牙螺纹Finished|精制的Finishing cement|水泥抹面Finishing coat|面漆Fire brick|耐火砖Fire door|防火门Fire extinguisher|灭火器Fire fighting truck|消防车Fire hose connection|消防软管接头Fire pump|消防水泵Fire safe type|耐火型Fire water|消防水Fire-proofing|防火层Fitting|管件fitting|管件Fitting to fitting|管件直接Fix point|固定点Fixed saddle|固定鞍座Flame arrester|阻火器Flame surface quenching|火焰表面淬火flame welding|气焊Flammable|可燃的；易燃的Flange|法兰Flange facing|法兰密封面；法兰面Flanged|法兰式的Flanged end|法兰端Flanged joint|法兰连接（接头）flap check valve|旋启式止回阀Flare|火炬Flare gas|火炬气Flaring test|扩口试验Flash drum|闪蒸罐Flash point|闪点Flat bar|扁钢Flat face|全平面；满平面Flat gasket|平垫片Flat metal gasket|平金属垫片Flat metal jacketed asbestos filled gasket|金属包石棉平垫片Flat nut|扁螺母Flat on bottom|底平Flat on top|顶平Flat ring gasket|环形平垫片Flat valve|盖阀Flat welding|平焊Flattening test|压扁试验Flexibility|柔性Flexibility characteristic|柔度特性Flexibility factor|柔性系数Flexibility stress|柔性应力Flexible solid wedge|挠性整体楔形闸板Flexible tube|挠性管Float trap|浮球式疏水阀Floating ball type|浮动球阀Floor|楼面Floor drain|地漏Flow diagram|流程图Flow meter|流量计Fluid|流体Fluid characteristics|流体特性Fluorescent penetrant inspection|荧光渗透检验Fluoroplastics|氟塑料Flush valve|冲洗阀Flush-bottom tank valve|罐底排污阀Flux|焊药（剂）Foam fire-fighting|泡沫消防Foam glass|泡沫玻璃Foam hydrant|泡沫栓Foam monitor|泡沫炮foam polystyrene|泡沫聚苯乙烯Foam station|泡沫站Foamed concrete|泡沫混凝土Foaming|发泡Foot|英尺Foot valve|底阀footing|基础Force|力Forged steel|锻钢Forged steel clevis|锻制U形夹Forged valve|锻造阀Forging|锻造的、锻造Foundation|基础foundation bolt|地脚螺栓Four-way plug valve|四通旋塞阀Fractionating tower|精馏塔Free|自由Free on board|船上交货,离岸价格Free on truck|敞车上交货Free to slide|自由滑动Free vibration|自由振动Freezing point|凝固点Frequency|频率Fuel gas|燃料气Fuel oil|燃料油Full bore|等径孔道full coupling|管接头full face|全平面；满平面Full jacketed|全夹套的full port|等径孔道Full thread|通长螺纹Full water test|盛水试验Funnel|漏斗Furan resin|呋喃树脂Furnace|炉子Furnace tube|炉管Fusion gas welding|气熔焊Future area|预留区Gage glass|玻璃液位计Galvanized iron|镀锌铁皮galvanized plain sheet|镀锌铁皮galvanized steel pipe|镀锌钢管Galvanized steel pipe|镀锌钢管Galvanized wire|镀锌铁丝Galvanized wire mesh|镀锌铁丝网Gamma radiography|γ射线照相gangway|走道；过道Gas analysis|气体分析Gas chromatograph|气象色谱仪Gas metal arc welding|金属极惰性气体保护电弧焊Gas turbine|燃气轮机Gas welding|气焊Gaseous corrosion|气相腐蚀Gas-holder|气柜Gasket|垫片Gas-shielded arc welding|气体保护电弧焊Gate valve|闸阀Gauze strainer|丝网粗滤器Gear pump|齿轮泵General carbon steel|普通碳素钢General plot plan|总图General structure low-alloy steel|普通低合金结构钢Generator|发电机gerritic alloy steel pipe|铁合金钢管Girder|桁架；主梁Gland|压盖Glass|玻璃Glass cloth|玻璃布Glass tube|玻璃管Glass wool|玻璃棉Globe type disc|球心型阀盘Globe valve|截止阀Gram|克Graphite phenolic plastics|石墨酚醛塑料Grate|篦子板grating|篦子板Gravel paving|碎石铺面Gravity settler|澄清器Grease injector|注油器Green|绿色的Grey|灰色的Grey cast iron|灰铸铁Groove|坡口Groove face|槽面Grooved metal gasket|槽形金属垫片Gross weight|毛重Ground level|地面Grounding|接地Grouting|灌浆；水泥砂浆填平Guide|导向架Gusset|角板；连接板Hair felt|发毡half coupling|半管接头Half coupling|半管接头Halogen gas leak test|卤气泄漏试验Hand lever|手柄Hand wheel|手轮Handhole|手孔Hand-operated valve|手动阀Hand-pump|手摇泵Handrail|栏杆Hanger|吊架Hard lead|硬铅Hard water|硬水Hardenability|可淬性Hardness test|硬度试验Harmonic analysis|谐振分析Hastelloy|耐蚀耐热镍基合金Hazardous area classification|危险区划分Hazardous area plan|危险区平面图Head room|净空Header|总管Header valve|总管阀Heat affected zone|热影响区Heat exchanger|换热器Heat resisting steel|耐热钢Heat treatment|热处理Heater|加热器Heating medium|载热体Heat-proof paint|耐热漆Heavy oil|重油Height|高度Helical gas-holder|螺旋式气柜Helical screw compressor|螺杆压缩机Hexagonal head bolt|六角头螺栓Hexagonal nut|六角螺母Hexagonal steel bar|六角钢High alloy steel|高合金钢High pressure|高压High pressure steam|高压蒸汽High silicon cast iron|高硅铸铁High strength steel|高强度钢High-carbon steel|高碳钢High-quality carbon steel|优质碳素钢Hinged expansion joint|带铰链膨胀节Hoisting beam|吊梁Hold|待定Homopolymer|均聚物Hook up drawing|连接图hoop stress|圆周应力Hopper|料斗Horizontal|水平的Horizontal installation|水平安装Hose connection|软管接头Hose reel|软管卷盘（筒）Hose station|软管站；公用工程站Hose valve|软管阀Hot insulation|保温Hot quenching|高温淬火Hot rolling|热轧Hot water|热水Hot working|热加工hot-rolling seamless pipe|热轧无缝钢管Hot-rolling seamless pipe|热轧无缝钢管Hot-water tracing|热水伴热Hour|（小）时Humidity|湿度Hydrant|消火栓Hydraulic operator|液压操纵器Hydraulic snubber|液压减振器Hydraulic test|水压试验Hydrogen|氢气Hydrogen embrittlement|氢脆I-beam|工字钢II-type support|II形管架II形管架|II-type supportImpact test|冲击试验Impact value|冲击值Impulse steam trap|脉冲式蒸汽疏水阀Inch|英寸Incinerator|焚烧炉Incoloy|耐热铬镍铁合金Incomplete fusion|未溶合Incomplete penetration|根部未焊透Inconel|铬镍铁合金inconel wire asbestos|带铬镍合金丝的石棉绳Index|索引；目录Indicator|指示器Induction hardening|感应（高频）硬化Industrial waste water|工业废水Inlet|入口Inline pump|管道泵Inner ring|内环Inorganic zinc-rich paint|无机富锌漆In-plane|面内Inquiry|询价inserted plate|预埋件Inside battery limit|项目区界内侧Inside diameter|内径Inside screw|内螺纹Inspection|检验Inspection hole|检查孔installation|安装Instrument air|仪气空气Instrument cable tray (duct)|仪表电缆槽（架）Instrument panel|仪表盘Instrumental analysis|仪器分析Insulation|隔热；融热层Insulation block|保温块Insulation break|隔热分界integral pipe flange|整体管法兰Integral pipe flange|整体管法兰Integral seat|整体（阀）座Inter cooler|中间冷却器Inter-department check|会签Intergranular corrosion|晶间腐蚀Interlock|联锁Intermittent welding|间断焊Internal approval plot plan|内部审查版设备布置图（“B”版）Internal force|内力Internal pressure stress|内压应力Intersection|相交Invert|管子内底Inverted bucket trap|倒吊桶式疏水阀Isolating valve|隔断阀Isometric drawing|轴测图Isothermal annealing|等温退火Isothermal quenching|等温淬火Issue|版次Itemized|编位号的Itemized equipment|编位号设备Jack screw|顶开螺栓；顶起螺栓Jacketed line|夹套管jacketed piping|夹套管Jacketed valve|夹套阀Jet pump|喷射泵Job No.|项目号Joule|焦耳Junction box|接线箱（盒）Key|键Key plan|分区索引图Kick-off meeting|开工会议Kieselguhr diatomite|硅藻土Killed steel|镇静钢Kilogram|千克(公斤)Kilopounds|千磅Kilowatts|千瓦Knock out drum|缓冲罐K形坡口|Double bevel groovelack of fusion|未溶合Ladder|直梯Lantern ring|笼式环lap|突缘短节Lap joint flange|松套法兰lap joint flange[LJF]|松套法兰Lap welding|搭焊Lapped joint|搭接接头；松套连接Lapped pipe end|翻边端Large end plain|大端为平的Large end thread|大端带螺纹latrolet|斜接支管台Latrolet|斜接支管台launching meeting|开工会议lavatory|盥洗室；厕所Lead|铅Leak test|泄漏试验Left hand thread|左螺纹Leg|支腿Legend|图例Length|长度Lens gasket|透镜式垫片Level gauge|液位计Lever|杠杆Lever and weight type|杠杆重锤式Lift check valve|升降式止回阀Lifting lug|吊耳Light|淡（浅）色的；轻的Lighting illumination|照明Lighting preventer|避雷针Limit of explosion|爆炸极限Limit rod|限制杆Limit stop|定值限位架Line list|管线表Line number|管线号line schedule|管线表Line spacing|管间距Line span|管道跨距Line-blind valve|管道盲板阀lined pipe|衬里管Lined pipe|衬里管Link butterfly valve|连杆式蝶阀Liquefied petroleum gas|液化石油气Liquid chromatograph|液相色谱仪Liquid expansion steam trap|液体膨胀式蒸汽疏水阀Liquid penetrant test|液体渗透检验List of nozzles|管口表Liter|升litre|升Live load|活荷载Load|荷载Load case|荷载工况Loading arm|装卸臂Local panel|就地盘Location|定位Lock nut|锁紧螺母Locked closed|在关闭状态下锁定Locked open|在开启状态下锁定Locker room|更衣室l0ng radius elbow|长半径弯头l0ng radius elbow|长半径弯头l0ng radius return|长半径180?弯头l0ng radius return|长半径180度弯头l0ngitudinal stress|纵向应力Loose float trap|自由浮球式疏水阀Loose hubbed flange|松套带颈法兰Loose plate flange|松套板式法兰Low alloy steel|低合金钢Low pressure|低压Low pressure steam|低压蒸汽Low-carbon steel|低碳钢L-type support|L形管架Lubricating oil|润滑油Lug|支耳；吊耳Lug type|凸耳式Lumped mass|集中质量L形管架|L-type supportMachine bolt|机螺栓；机螺钉Magnesium carbonate|碳酸镁Magnetic particle test|磁粉探伤Maintenance room|维修间Make-up|补充Male face|凸面Malleable iron|可锻铸铁Man-day|工日Manganese bronze|锰青铜Manhole|人孔Man-hour|工时manifold|总管Man-month|人月Manometer|压力计Manual and automatic inert gas tungsten arc welding|手工或自动惰性气体保护钨极电弧焊manually operated valve|手动阀Manufacture|制造者,制造厂Martensitic stainless steel|马氏体不锈钢Martin steel|平炉钢（马丁钢）Mass spectrometric analysis|质谱分析Mastic weatherproof coating|玛帝脂保护层Match line|接续线Material specification break|等级分界Material status report|材料情况报告Material take-off|材料统计Maximum|最大Measuring tank|计量槽Mechanical trap|机械式疏水阀Mechanical vibration|机械振动Medium alloy steel|中合金钢Medium pressure|中压Medium pressure steam|中压蒸汽Medium-carbon steel|中碳钢Melter|融解槽Melting point|融点Member|构件;元件Metal expansion steam trap|金属膨胀式蒸汽疏水阀Metal hose|金属软管metal jacketing|金属保护层Metallic stuffing|金属填料Meter|米Metering pump|计量泵Methanator|甲烷化器Metric thread|公制螺纹Micro crack|微裂纹Microscopic test|金相试验Millimeter|毫米Million pascal|百万帕斯卡Mineral wool|矿（渣）棉Minimum|最小Minute|分Mismatch|错位mitre bend|斜接弯管Mitre bend|斜接弯管Mixer|混合器Mixing valve|混合阀Model|模型Modulus of elasticity|弹性模量Moisture-proof packing|防潮湿包装Molecular sieve|分子筛Molybdenum steel|钼钢Moment|力矩Moment of inertia|惯性矩Monel|蒙乃尔（注：镍及铜合金）Monomer|单体Most frequent wind direction|主导风向Mother liquor|母液Motor|电动机Motor hoist|电动葫芦Multiple below|多波Multiple stages compressor|多级压缩机Multiport valve|多通路阀Naphtha|石脑油Natural frequency|固有频率Natural frequency mode|固有振动型式Natural gas|天然气Natural rubber|天然橡胶Natural white rubber gasket|天然白橡胶垫片Needle valve|针阀Neoprene|氯丁橡胶Net positive suction head|净正吸入压头Net weight|净重Network of pipes|管网Neutralization tank|中和池Newton|牛顿Nickel copper alloy|镍铜合金Nickel steel|镍钢Nipolet|短管支管台nipolet|短管支管台Nipple|短节Nitric acid|硝酸Nitridation|渗氮Nitrile butadiene rubber|丁腈橡胶Nitrogen|氮气Node|节点Node number|节点号Nodular cast iron|球墨铸铁nodular graphite iron|球墨铸铁Noise level|噪声值Nomenclature|命名表；管道表Nominal diameter|公称直径Nominal pressure|公称压力Non graphited compressed white asbestos gasket|无石墨压缩白石棉垫片Non-destructive testing|无损检验Nonferrous alloy|非铁合金Non-ferrous metal|有色金属Non-itemized equipment|无位号设备Non-metallic gasket|非金属垫片Non-metallic material|非金属材料non-return valve|切断式止回阀Non-rising stem|非升杆式（暗杆）Non-rotary valve|非旋转式阀Non-slam check valve|无撞击声止回阀Normalizing|正火Normally closed|常闭Normally open|常开North|北Nozzle|管口Nozzle orientation|管口方位Nut|螺母Nylon plastic|尼龙塑料Observation door|观察孔Octagonal ring gasket|八角环形垫片Off site|界外Office|办公室offset bend|偏置弯管Offset bend|偏置弯管Offset disc butterfly valve|偏心阀板蝶阀Ohm|欧（姆）Oil quenching|油淬火On stream|运行中One end thread|一端带螺纹Open bucket trap|浮桶式疏水阀open top bucket trap|浮桶式疏水阀open-steel|沸腾钢Operating pressure|工作压力Operating temperature|工作温度Opposite|相反（的）；对面（的）Orange colour|橙色的Organic silicon paint|有机硅漆Orientation|方位Orifice|孔板orifice flange|孔板法兰Orifice flange|孔板法兰Origin of coordinate|坐标原点O-ring|O形环Outer ring|外环；外定位环Outlet|出口Out-plane|面外Outside diameter|外径Outside screw & yoke (OS & Y)|外螺纹阀杆及阀轭Oval ring gasket|椭圆环形垫片Over haul|大修Over-dimension cargo|超尺寸运输Overhead welding|仰焊Overlap|焊瘤Over-sea mean level|海平面标高Owner|业主Owner approval plot plan|用户审查版设备布置图（“C”版）Oxygen|氧气O形环|O-ringPackage unit|成套设备Packing|包装Packing box|填料箱；填料函Packing list|装箱单Pad type flange|盘座式法兰Painting|涂漆Panel|盘（操作盘）Parallel|平行；平行的Part number|件号Partition wall|隔墙Parts per million|百万分之一Paving area|铺砌区Peak stress|峰值应力peep door|观察孔Pelletizer|造粒机Percentage el0ngation|延伸率Period|周期Perlite|珍珠岩Permanent filter|永久过滤器Perpendicular|垂直；正交；垂直的Personal protection|人身保护Phase|相位Phenolic paint|酚醛漆Phosphor bronze|磷青铜Pig iron|生铁Pile|桩Pilot operated safety valve|引导阀操纵的安全泄气阀Pin|销Pin hole|针孔。
中科大_盲信号处理_第4章
(4-23)
Q D1/2 UT s
利用(4-24)式获得的 Q ，容易验证
(4-24)
E[Qy (t )y T (t )QT ] Q[R x (0) 2 I M ]QT I N
2 2
3）对于 i, j 1, 2,, N ， i j ， i (t ) / j (t ) 是时变的。
2 2
2. 分离准则与代价函数
在上述 3 个假设下，Matsuok 等人已经证明下列的一个重要结论： 设计一个分离矩阵 W ， y (t ) Wx(t ) WHs(t ) ，如果对于任意一个时刻 t ，都有
球化后的输出为
(4-25)
z (t ) Qx(t ) QHs(t ) Qn(t ) Vs(t ) Qn(t )
3. 利用单延时的 R z ( ) 估计正交矩阵 V 根据(4-24)式，有
(4-26)
0 ， R z ( ) VR s ( )VT
注意到 R z ( ) 是对称矩阵， R s ( ) 是对角矩阵。
(4-21)
其中， U s 是 M N 的矩阵， U n 是 M ( M N ) 的矩阵， [ U s U n ] 是正交矩阵， 是对 角矩阵，且
2 2 2 2 diag ( 12 , 2 ,, N , N 1 , , M )
(4-22)
噪声方差的估计为
2 2 ˆ 2 ( N 1 M ) / ( M N )
(4-3)
Rsi ( ) E[ si (t ) si (t )] E[ si (t )]E[ si (t )] 0
Procedings of the IASTED International Conference APPLIED SIMULATION AND MODELLING
Procedings of the IASTED International ConferenceAPPLIED SIMULATION AND MODELLINGSeptember3-5,2003,Marbella,SpainA time-frequency approach to blind separation of under-determinedmixture of sourcesA.MANSOURLab.E I,ENSIETA,29806Brest cedex09,(FRANCE).mansour@M.KAW AMOTODept.of Electronic andControl Systems Eng.,Shimane University,Shimane690-8504,(JAPAN)kawa@ecs.shimane-u.ac.jpC.PuntonetDepartamento de Arquitectura yTecnologia de computadores,Universidad de Granada,18071Granada,(SPAIN).carlos@atc.ugr.esABSTRACTThis paper deals with the problem of blind separation of under-determined or over-complete mixtures(i.e.more sources than sensors).Atfirst a global scheme to sepa-rate under-determined mixtures is presented.Then a new approach based on time-frequency representations(TFR) is discussed.Finally,some experiments are conducted and some experimental results are given.KEY WORDSICA,BSS,Time-Frequency domain,over-complete or under-determined mixtures1.IntroductionBlind separation of sources problem is a recent and an im-portant signal processing problem.This problem involves recovering unknown sources by only observing some mixed signals of them[1].Generally,researchers assume that the sources are statistically independent from each other and at most one of them can be a Gaussian signal[2]. Other assumptions can be also founded in the literature concerning the nature of the transmission channel(i.e.an instantaneous or a memoryless channel,a convolutive or a memory channel,and a non-linear channel).In addition,a widely used assumption considers that the number of sen-sors should be equal or greater(for subspace approaches) than the number of sources.These assumptions are fairly satisfied in many divers applications such as robotics, telecommunication,biomedical engineering,radars,etc., see[3].In recent applications linked to special scenarios in telecommunication(as satellite communication in double-talk mode),robotics(for exemple,robots which imitate human behavior)or radar(in ELectronic INTelli-gence”ELINT”applications),the assumption about the number of sensors can not be satisfied.In fact,in the latter applications the number of sensors is less than the number of sources and often we should deal with a mono-sensor system with two or more sources.Recently,few authors have considered the under-determined mixtures.Thus by using overcomplete repre-sentations,Lewicki and Sejnowski in[4]present an algo-rithms to learn overcomplete basis.Their algorithm uses a Gaussian approximation of probability density function (PDF)to maximize the probability of the data given the model.Their approach can be considered as a generaliza-tion of the Independent Component Analysis(ICA)[2]in the case of instantaneous mixtures.However,in this ap-proach,the sources should be sparse enough to get good ex-perimental results,otherwise the sources are being mapped down to a smaller subspace and there is necessary a loss of ing the previous approach,Lee et al.[5] separate successfully three speech signals using two micro-phones.On the other hand,When the sources are sparsely distributed,at any time t,at most one of sources could be significantly different from zero.In this case,estimating the mixing matrix[6,7,8]consists offinding the direc-tions of maximum data density by simple clustering ing Reimannian metrics and Lie group structures on the manifolds of over-complet mixture matrices,Zhang et al.[9]present a theoretical approach and develop an al-gorithm which can be considered as a generalization of the one presented in[10].The algorithm of Zhang et al.up-date the weight matrix by minimizing a kullback-Leibler divergence by using natural learning algorithm[11].In the general case,one can consider that separation of over-complete mixtures still a real challenge for the sci-entific community.However,some algorithms have been proposed to deal with particular applications.Thus for bi-nary signals used in digital communication,Diamantaras and Chassioti[12,13]propose an algorithm based on the PDF of the observed mixed signals.The pdf of the ob-servation signals have been modeled by Gaussian pdf and estimated from the histogram of the observed -ing differential correlation function,Deville and Savoldelli [14]propose an algorithm to separate two sources from noisy convolutive mixtures.The proposed approach re-quires the sources to be long-term non-stationary signals and the noise should be long-terme stationary ones.The previous statement means that the sources(resp.noise) should have different(resp.identically)second order statis-tics at different instances separated by a long period.2.Channel ModelHereinafter,we consider that the sources are non Gaus-sian signals and statistically independent from each other.In addition,we assume that the noise is an additive whiteGaussian noise (AWGN).Letdenote the source vector at any time t,is mixing vector and is a AWGN vector.The channel is represented by a full rank real and constant matrix ().H ( )Channel+B(t)S(t)Figure 1.General structure.The separation is considered achieved when the sources are estimated up to a scale factor and a permuta-tion.That means the global matrix can be written as:here,is a weight matrix,is a permutation matrix and is a non-zero diagonal matrix.For a sake of simplic-ity and without loss of generality,we will consider in the following that:Where is an invertible matrix and is a full rankrectangular matrix.3.A Separation SchemeIn the case of over-complete mixtures (),the invert-ibility of the mixing matrix becomes an ill-conditioned problem.That means the Independent Component Analy-sis (ICA)will be reduced to extract independent signals which are not necessarily the origine sources,i.e.the sep-aration can not give a unique solution.Therefore,further assumptions should be considered and in consequence suit-able algorithms could be developed.Thus,two strategies can be considered:At first one can identify the mixing matrix then us-ing this estimated matrix along with important infor-mation about the nature or the distributions of the sources,we should retrieve the original sources.In many applications (such as speech signals,telecom-munications,etc ),one can assume the sources havespecial features (constant modulus,frequency prop-erties,etc ).Using sources’specifics,the separation becomes possible in the classic manner,i.e.up to per-mutation and a scale factor.Beside the algorithms cited and discussed in the intro-duction of our manuscript,few more algorithms can be founded in the literature.The latter publications are dis-cussed in this section.3.1Identification &SeparationOne of the first publications on the identification of under-determined mixtures was proposed by Cardoso [15].In his manuscript,Cardoso proposed an algorithm based only on fourth-order cumulant.In fact,using the symmetries of quadricovariance tensor,an identification method based on the decomposition of the quadricovariance was proposed.Recently,Comon [16]proved using an algebraic approach,that the identification of static MIMO (Multiple Inputs Multiple Outputs)with fewer outputs than inputs is possible.In other words,he proved that the CANonical Decomposition (CAND)of a fourth-order cross-cumulant tensor can be considered to achieve the identification.In addition,he proved that ICA is a symmetric version of ing a Sylveter’s theorem in multilinear algebra and the fourth order cross cumulant tensor,he proposed an algorithm to identify the mixing matrix in the general case.To recover d-psk sources,comon proposes alsoa non-linear inversion ofby adding some non-linear equations and using the fact that the d-psk signals satisfyspecial polynomial properties (i.e.).Later on,Comon and Grellier [17]proposed an extension of the previous algorithm to deal with different communication signals (MSK,QPSK and QAM4).Similar approach was also proposed by De Lathauwer et al.,see [18].Finally,Taleb [19]proposes a blind identification al-gorithm of M-inputs and 2-outputs channel.He proved thatthe coefficients of the mixing matrixare the roots of a polynomial equations based on the derivative of the sec-ond characteristic function of the observed signals.The uniqueness of the solution is proved using Darmois’Theo-rem [20].3.2Direct SeparationHere,we discuss methods to separate special signals.As it is mentioned in the previous subsection that Comon et al.[16,17]proposed an algorithm to separate communication signals.Nakadai et al.[21,22]addressed the problem of a blind separation of three mixed speech signals with the help of two microphones by integrating auditory and vi-sual processing in real world robot audition systems.Theirapproach is based on direction pass-filters which are imple-mented using the interaural phase difference and the inter-aural intensity difference in each sub-band -ing Dempster-Shafer theory,they determine the direction of each sub-band frequency.Finally,the waveform of one sound can be obtained by a simple inverse FFT applied to the addition of the sub-band frequencies issued by the spe-cific direction of that speaker.Their global system can per-form sound source localization,separation and recognition by using audio-visual integration with active movements.4.Time-Frequency ApproachThe algorithm proposed in this section is based on time-frequency distributions of the observed signals.To our knowledge,few time-frequency methods have been devoted to the blind separation of MIMO channel.In fact,for MIMO channel with more sensors than sources, Belouchrani and Moeness[23]proposed a time-frequency separation method exploiting the difference in the time-frequency signatures of the sources which are assumed to be nonstationary multi-variate process.Their idea consists on achieving a joint diagonalization of a combined set of spatial time-frequency distributions which have been defined in their paper.It is clear from the discussion of the previous sections that the identification of MIMO channel is possible.How-ever,the separation is not evident in the general case.The few published algorithms for the under-determined matter are very linked to signal features of theirs applications.In our applications,an instantaneous static under-determined mixture of speech signals is considered.This problem can be divided into two steps:Atfirst an identification algorithm should be applied.For the moment,we didn’t develop a specific identi-fication algorithm.Therefore,any identification algo-rithms previously mentioned can be used.Let us assume that the coefficient of the mixing matrixhave been estimated.The question becomes How can we recover the sources from fewer sensors?To answer this question,we consider in this section the separation of a few speech signals(for the instance, we are considering just two or at most three sources) using the output of a single microphone(i.e.Multiple Inputs Single output,MISO channel).Recently,time-frequency representations(TFR) have been developed by many researchers[24]and they are considered as very powerful signal processing tools.In the literature,many different TFR have been developed as Wigner-Ville,Pseudo-Wigner-Ville,Smooth Pseudo-Wigner-Ville,Cho-Willims,Born-Jordan,etc.In a previous study[25],we found that for simplicity and performance reasons,the Pseudo-Wigner-Ville can be considered as a good TFR candidate.Here we present a new algorithm based on time-frequency representations of the observed signals(TFR)to separate a MISO channel with speech sources.It is known that speech signals are non-stationary signals.However within phonemes(about80ms of duration)the statistics of the signal are relatively constant[26].On the other hand,It is well known that voiced speech are quasi-periodic signals and the non-voiced signals can be considered as white filtered noise[27].Within a small window corresponding to51ms,the pitch can be slightly change.Therefore,one can use this property to pick up the frequency segments of a speaker.The pitch can be estimated using divers techniques[28].Using the previous facts and Pseudo-Wigner-Ville representations,one can separate up to three speech signals from one observed mixed signal of them.To achieve that goal,we assume that the time-frequency signatures of the sources are disjoints.Atfirst,one should calculate the TFR of the observed signal.Then,in the time-frequency space, we plot a regular grilled.The dimensions of the a small cell of the grilled are evaluated based on the properties of the speech signals and the sampling frequency.Therefore, these dimensions can be considered as10to20ms in length (i.e.time axis)and5to10%of the sampling frequency value in the vertical axis.Once we plot the grilled,we estimate the energy average in each cell and a threshold is applied to distinguish noisy cells from other.Then the cell with the maximum energy is considered as a potential pitch of one speaker and it is pointed out.After that,we merge in a set of cells,all cells with high level of energy in the neighborhood of the previous cell.At least one har-monic of the pitch should be also selected.The previous steps should repeated as necessary.Finally,the obtained map can be considered as a bi-dimensional time-frequency filters which should be applied on the mixed -ing a simple correlation maximization algorithm,one can find the different pieces corresponding to the speech of one speaker.5.Experimental ResultsTo demonstrate the validity of the proposed algorithm men-tioned in section4,many computer simulations were con-ducted.Some results are shown in this section.We consid-ered the following two-input and one-output system.(1)The sources were male and female voices which were recorded by8[KHz]sampling fre-quency.The TFR was calculated by using128data of the observed signal.Figure2shows the results obtained by applying the proposed algorithm(last paragraph in section4)to the ob-served signal.From thisfigure,one might think that the estimated signals are different from the original signals. However,if one hear the estimated signals,one can see that the two original sources and are separated from the observed signal by the proposed algorithm.6.ConclusionThis paper deals with the problem of blind separation of under-determined(or over-complete)mixtures(i.e.more sources than sensors).Atfirst,a survey on blind separation algorithms for under-determined mixtures is given.A sep-aration scheme based on identification or direct separation is discussed.A new time-frequency algorithm to separate speech signals has been proposed.Finally,some experi-ments have been conducted and the some experimental re-sults are given.Actually,we are working on a project con-cern the separation of under-determined mixtures.Further results will be the subject of future communications.References[1]A.Mansour, A.Kardec Barros,and N.Ohnishi,“Blind separation of sources:Methods,assumptions and applications.,”IEICE Transactions on Funda-mentals of Electronics,Communications and Com-puter Sciences,vol.E83-A,no.8,pp.1498–1512, August2000.[2]on,“Independent component analysis,a newconcept?,”Signal Processing,vol.36,no.3,pp.287–314,April1994.[3]A.Mansour and M.Kawamoto,“Ica papers classi-fied according to their applications&performances.,”IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences,vol.E86-A,no.3,pp.620–633,March2003.[4]M.Lewicki and T.J.Sejnowski,“Learning non-linear overcomplete representations for efficient cod-ing,”Advances in neural Information Processing Sys-tems,vol.10,pp.815–821,1998.[5]T.W.Lee,M.S.Lewicki,M.Girolami,and T.J.Se-jnowski,“Blind source separation of more sources than mixtures using overcomplete representations,”IEEE Signal Processing Letters,vol.6,no.4,pp.87–90,April1999.[6]P.Bofill and M.Zibulevsky,“Blind separation ofmore sources than mixtures using sparsity of their short-time fourier transform,”in International Work-shop on Independent Component Analysis and blind Signal Separation,Helsinki,Finland,19-22June 2000,pp.87–92.[7]P.Bofill and M.Zibulevsky,“Underdetermined blindsource separation using sparse representations,”Sig-nal Processing,vol.81,pp.2353–2363,2001. [8]P.Bofill,“Undetermined blind separation of delayedsound sources in the frequency domain,”NeuroCom-puting,p.To appear,2002.[9]L.Q.Zhang,S.I.Amari,and A.Cichocki,“Nat-ural gradient approach to blind separation of over-and under-complete mixtures,”in First International Workshop on Independent Component Analysis and signal Separation(ICA99),J.F.Cardoso,Ch.Jutten, and Ph.loubaton,Eds.,Aussois,France,11-15Jan-uary1999,pp.455–460.[10]M.Lewicki and T.J.Sejnowski,“Learning overcom-plete representations,”Neural Computation,vol.12, no.2,pp.337–365,2000.[11]S.I.Amari,A.Cichocki,and H.H.Yang,“A newlearning algorithm for blind signal separation,”in Neural Information Processing System8,Eds.D.S.Toureyzky et.al.,1995,pp.757–763.[12]K.Diamantaras and E.Chassioti,“Blind separationof n binary sources from one observation:A deter-ministic approach,”in International Workshop on In-dependent Component Analysis and blind Signal Sep-aration,Helsinki,Finland,19-22June2000,pp.93–98.[13]K.Diamantaras,“Blind separation of multiple binarysources using a single linear mixture,”in Proceed-ings of International Conference on Acoustics Speech and Signal Processing2001,ICASSP2000,Istanbul, Turkey,Jun2000,pp.2889–2892.[14]Y.Deville and S.Savoldelli,“A second order dif-ferential approach for underdetermined convolutive source separation,”in Proceedings of International Conference on Acoustics Speech and Signal Process-ing2001,ICASSP2001,Salt Lake City,Utah,USA, May7-112001.[15]J.F.Cardoso,“Super-symetric decomposition of thefourth-order cumulant tensor.blind identification of more sources than sensors.,”in Proceedings of Inter-national Conference on Speech and Signal Process-ing1991,ICASSP’91,Toronto-Canada,May1991, pp.3109–3112.[16]on,“Blind channel identification and extrac-tion of more sources than sensors,”in In SPIE Confer-ence on Advanced Algorithms and Architectures for Signal Processing,San Diego(CA),USA,July19-24 1998,pp.2–13,Keynote address.[17]on and O.Grellier,“Non-linear inversion ofunderdetermined mixtures,”in First International Workshop on Independent Component Analysis andFigure2.Simulations Results:(a)Source signal(b)Source signal(c)Observed signal(d)Estimated signal of(e)Estimated signal ofsignal Separation(ICA99),J.F.Cardoso,Ch.Jut-ten,and Ph.loubaton,Eds.,Aussois,FRANCE,11-15 January1999,pp.461–465.[18]L.De Lathauwer,on,B.De Moor,and J.Van-dewalle,“ICA algorithms for3sources and2sen-sors,”in IEEE SP Int Workshop on High Order Statis-tics,HOS99,Caeserea,Israel,12-14June1999,pp.116–120.[19]A.Taleb,“An algorithm for the blind identification ofn independent signals with2sensors,”in Sixth Inter-national Symposium on Signal Processing and its Ap-plications(ISSPA2001),M.Deriche,Boashash,and W.W.Boles,Eds.,Kuala-Lampur,Malaysia,August 13-162001.[20]G.Darmois,“Analyse g´e n´e rale des liaisons stochas-tiques,”Rev.Inst.Intern.Stat.,vol.21,pp.2–8,1953.[21]K.I.Nakadai,K.Hidai,H.G.Okuno,and H.ki-tano,“Real-time speaker localization and speech sep-aration by audio-visual integration,”in17th inter-national Joint Conference on Artificial Intelligence (IJCAI-01),Seatle,USA,August2001,pp.1425–1432.[22]H.G.Okuno,K.Nakadai,T.Lourens,and H.kitano,“Separating three simultaneous speeches with two microphones by integrating auditory and visual pro-cessing,”in European Conference on Speech Process-ing,Aalborg,Denmark,September2001,pp.2643–2646.[23]A.Belouchrani and M.G.Amin,“Blind source sep-aration based on time-frequency signal representa-tions,”IEEE Trans.on Signal Processing,vol.46, no.11,pp.2888–2897,1998.[24]P.Flandrin,Time-Frequency/Time-Scale analysis,Academic Press,Paris,1999.[25]D.Le Guen and A.Mansour,“Automatic recogni-tion algorithm for digitally modulated signals,”in6th Baiona workshop on signal processing in communi-cations,Baiona,Spain,25-28June2003,p.To ap-pear.[26]J.Thiemann,Acoustic noise suppression for speechsignals using auditory masking effects,Ph.D.thesis, Department of Electrical&Computer Engineering, McGill University,Canada,July2001.[27]R.Le Bouquin,Traitemnet pour la reduction du bruitsur la parole application aux communications radio-mobiles.,Ph.D.thesis,L’universit´e de Rennes I,July 1991.[28]A.Jefremov and B.Kleijn,“Sline-based continuous-time pitch estimation,”in Proceedings of Interna-tional Conference on Acoustics Speech and Signal Processing2002,ICASSP2002,Orlando,Florida, U.S.A,13-17May2002.。
生物阻抗胃动力信号采集方法
生物阻抗胃动力信号采集方法*李章勇1,2，魏进民1，任超世2，沙洪2，王伟11.重庆邮电大学生物信息学院，重庆（400065）2.中国医学科学院、中国协和医科大学生物医学工程研究所，天津（300192）E-mail：li9547@摘要：临床将胃电信号作为胃动力紊乱诊断依据，但胃电信号并不完全对应于胃的活动，特别是胃的蠕动、胃的排空等机械活动。
如果将与胃机械活动对应的胃阻抗信息也作为诊断依据之一，那就为建立新的、更准确的胃动力评价方法提供可能。
这也是生物阻抗技术在胃功能评价方面的最新发展。
本文基于阻抗法设计了一个新的获取胃动力信息的系统，通过计算机串口控件收发数据，采集到了阻抗和同步胃电信号。
本系统不仅获得常规检测参数胃电，也通过小波变换技术分离出与胃电同步的胃阻抗信息，为临床胃动力学研究提供新途径。
关键词：胃动力，生物阻抗，信号处理，小波变换1. 导言胃肠疾病是常见病，多发病，其发生率超过总人口的1%。
它危害人民健康，严重影响了人们的工作、学习和生活质量。
在我国，胃肠病例中的50%与胃动力异常相关,已受到消化内科医生的特别关注和重视[1]。
胃肠动力学是一门正在迅速发展的、多学科交叉的新兴学科。
在国外，胃动力学的研究也还是一个十分年轻的医学前沿课题，国内则起步较晚。
长期以来，人们对胃动力功能的研究远远落后于对胃的内、外分泌功能及胃的形态学的研究。
其中一个很重要的原因就是缺乏方便、有效的胃动力学检查方法或手段[2]。
在国内外现有的一些检测方法中，腔内压测量、恒压器检查等为创伤性方法，患者难于接受；胃排空闪烁显像、放射性核素呼气试验、不透X线标志物法胃肠道通过时间检查等要使用核素或射线，对患者有害，不宜长时间、多次重复使用；超声方法虽然可以观察到胃排空或胃运动情况，但要用于消化过程的长时间检查和评价，在操作和技术上还存在不少困难，难于实现；体表胃电图无创、方便，但胃电只反映胃的电活动频率，与胃动力，特别是胃运动的相关性不强。
Blind source separation utilizing a spatial fourth
专利名称：Blind source separation utilizing a spatialfourth order cumulant matrix pencil发明人：John F. Dishman,G. Patrick Martin,Edward R.Beadle申请号：US10360631申请日：20030210公开号：US20030204380A1公开日：20031030专利内容由知识产权出版社提供专利附图：摘要：Blind source separation (BSS) of statistically independent signals with lowsignal-to-noise plus interference ratios under a narrowband assumption utilizescumulants in conjunction with spectral estimation of the signal subspace to perform the blind separation. The BSS technique utilizes a higher-order statistical method, specifically fourth-order cumulants, with the generalized eigen analysis of a matrix-pencil to blindly separate a linear mixture of unknown, statistically independent, stationary narrowband signals at a low signal-to-noise plus interference ratio having the capability to separate signals in spatially and/or temporally correlated Gaussian noise. This BSS provides the ability to blindly separate signals in situations where no second-order technique has been found to perform the blind separation satisfactorily, for example, at a low signal-to-noise ratio when the number of sources equals the number of sensors or when the noise is spatially and temporally colored.申请人：DISHMAN JOHN F.,MARTIN G. PATRICK,BEADLE EDWARD R.更多信息请下载全文后查看。
基于广义周期性的单通道多分量正弦调频信号分离和参数估计
基于广义周期性的单通道多分量正弦调频信号分离和参数估计朱航;张淑宁;赵惠昌【摘要】针对噪声条件下的单通道多分量正弦调频(SFM)信号，该文提出一种信号分离和参数提取方法。
利用正弦调频信号的广义周期性进行奇异值分解，以求出分量信号的调制频率；通过离散点搜索，估计出分量信号的调频(FM)初始相位、调制指数及载频，并对这些估计值利用信赖域算法进行优化，减小误差；利用内积计算，估计分量信号的幅度和初始相位。
此外，还利用自相关矩阵特征值分解估计混合信号的信噪比(SNR)，并根据信噪比确定停止分解的阈值。
在仿真与分析中，针对具体的信号详细说明了该方法的各步骤，并在不同信噪比条件下分析了该方法的参数估计精确度。
%A method of single channel source separation and parameters estimation of multi-component Sinusoid Frequency Modulation (SFM) signal is proposed, the generalized periodic is derived and it can be used to estimate the modulation frequency by singular value decomposition. The estimated modulation frequency can be used as a known quantity, then by searching the discrete points and optimization calculating with the trust region algorithm, the Frequency Modulation (FM) initial phase, the FM index and the carrier frequency are determined. Finally, the amplitude and the initial phase can be determined by calculating the inner product. In addition, the proposed method can estimate the Signal Noise Ratio (SNR) by using eigenvalue decomposition, and the SNR can be used to determine the threshold adaptively. In the simulation, the method is demonstrated by separating a specific signal, and it is supposed that the proposed method is effective in different SNRs.【期刊名称】《电子与信息学报》【年(卷),期】2014(000)010【总页数】7页(P2438-2444)【关键词】信号处理;多分量正弦调频信号;信号分离;参数估计;广义周期性;自适应阈值【作者】朱航;张淑宁;赵惠昌【作者单位】南京理工大学电子工程与光电技术学院南京 210094; 解放军73015部队湖州 313000;南京理工大学电子工程与光电技术学院南京 210094;南京理工大学电子工程与光电技术学院南京 210094【正文语种】中文【中图分类】TN971.1作为非线性调频信号特例的正弦调频(Sinusoid Frequency Modulation, SFM)信号由于具有频率时变、截获概率低等特点，在雷达目标检测、引信抗干扰等领域得到广泛应用。
密度梯度离心法分离脐血干细胞：分离介质的筛选★
密度梯度离心法分离脐血干细胞：分离介质的筛选★郭继强;刘爱兵;王东平;王黎明【期刊名称】《中国组织工程研究》【年(卷),期】2013(000)023【摘要】背景：应用脐血分离干细胞的目的是获得以干细胞为主要群体的单个核细胞群，密度梯度离心法是最简单有效的方法之一。
密度梯度离心法使用的分离介质以聚蔗糖泛影葡胺最为常用，但哪种浓度获得干细胞最多，目前尚未深入研究。
目的：探讨密度梯度离心法分离人脐血干细胞分离介质的最佳浓度，建立临床级干细胞分离应用方案。
方法：采用两步法分离脐血流程，先用羟乙基淀粉沉淀脐血红细胞，再使用质量浓度分别为(1.0730±0.0001)，(1.0750±0.0001)，(1.0770±0.0001) g/mL的分离液，分离沉淀脐血红细胞后的上清液，得到单个核细胞，分别计数细胞获得率及细胞存活率。
采用流式细胞仪测定单个核细胞表面标志物，将各亚组分绘制成直方图或散点图，分析所得脐血单个核细胞中所含单个核细胞亚组分的比例和绝对数量。
结果与结论：应用质量浓度(1.0730±0.0001)g/mL的分离液可得到最大比例间充质干细胞群，是分离间充质干细胞的最佳质量浓度。
使用质量浓度(1.0750±0.0001) g/mL的分离液可得到较高比例的造血干细胞群，是分离造血干细胞的最佳质量浓度。
使用质量浓度(1.0770±0.0001) g/mL 的分离液得到细胞总数最高，但获得的间充质干细胞、造血干细胞比例最低。
采用两步法分离干细胞流程，建立严格的实验室条件和标准，可获得密度梯度离心法分离人脐血干细胞的最佳分离方案。
% BACKGROUND: The purpose for isolating stem cel s from the cord blood is to obtain the mononuclear cel populations with the stem cel s as the major group, and density gradientcentrifugation is one of the simplest and most effective ways. Polysucrose diatrizoate is the most commonly used separation medium for density gradient centrifugation, but there is no in-depth research on which method can obtain more stem cel s. OBJECTIVE: To investigate the optimal density of separation medium for isolating human umbilical cord blood stem cel s using density gradient centrifugation method, and to establish the isolation method of stem cel s for clinical application. METHODS: Umbilical cord blood was separated with two-step method. Firstly, hydroxyethyl starch was used to sediment cord blood erythrocyte, and the suspension was obtained, then the separation media with three kinds of concentrations (1.073 0±0.000 1), (1.075 0±0.000 1) and (1.077 0±0.000 1) g/mL was used to isolate the suspension and the mononuclear cel s were obtained. The harvest rate and the survival rate of the cel s were counted. Their surface markers of mononuclear cel s were identified with flow cytometry; histogram or scatter plot of each subset was plotted to analyze their proportions and absolute numbers of sub-cel populations. RESULTS AND CONCLUSION: The separation medium with the concentration of (1.073 0±0.000 1) g/mL could obtain the mesenchymal stem cel populations with the largest proportion, which was the optimal density for isolating the mesenchymal stem cel s. The separation medium with the concentration of (1.075 0± 0.000 1) g/mL could obtain the hemopoietic stem cel populations with the larger proportion, which was the optimal density for isolating the hemopoietic stem cel s. The separation medium with the concentration of (1.077 0±0.000 1) g/mL could obtain the largestnumber of stem cel s, but the obtained mesenchymal stem cel s and hematopoietic stem cel s had the lowest proportion. Application of two-step method for isolating stem cel s and the establishment of strict laboratory conditions and standards are the optimal programs for isolating the human umbilical cord blood stem cel s using density gradient centrifugation method.【总页数】7页(P4189-4195)【作者】郭继强;刘爱兵;王东平;王黎明【作者单位】武警总医院医学实验中心，北京市 100039;武警总医院医学实验中心，北京市 100039;武警总医院医学实验中心，北京市 100039;武警总医院医学实验中心，北京市 100039【正文语种】中文【中图分类】R394.2【相关文献】1.密度梯度离心法分离脐血干细胞：分离介质的筛选 [J], 郭继强;刘爱兵;王东平;王黎明;2.犬骨髓间充质干细胞分离纯化和成骨诱导分化:Ficoll液密度梯度离心法体外分离的可行性 [J], 解芳;滕利;蔡磊;徐家杰;靳小雷;肖苒;曹谊林3.密度梯度离心法结合贴壁筛选法分离培养大鼠骨髓间充质干细胞的生物学特性[J], 聂文波;汪明星;张振华;孙付杰4.Percoll不连续密度梯度离心法分离红螯光壳螯虾血细胞 [J], 傅蓉蓉;李钫;杨丰5.密度梯度离心法分离单采浓缩白细胞悬液中单个核细胞方法探讨 [J], 王富英; 廖群艳因版权原因，仅展示原文概要，查看原文内容请购买。
- 1、下载文档前请自行甄别文档内容的完整性，平台不提供额外的编辑、内容补充、找答案等附加服务。
- 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
- 3、如文档侵犯您的权益，请联系客服反馈,我们会尽快为您处理(人工客服工作时间：9:00-18:30)。
BLIND SEPARATION USING ABSOLUTE MOMENTS BASED ADAPTIVE ESTIMATINGFUNCTIONJuha Karvanen and Visa KoivunenSignal Processing LaboratoryHelsinki University of TechnologyP.O.Box3000,FIN-02015HUT,FinlandTel.+35894512455Fax+3589460224juha.karvanen,visa.koivunen@hut.fiABSTRACTWe propose new absolute moment based estimating func-tions for blind source separation purposes.Absolute mo-ments are a computationally simple choice that can alsoadapt to the skewness of source distributions.They havelower sample variance than cumulants employed in manywidely used ICA(Independent Component Analysis)meth-ods.The complete estimating function consists of two partsthat are sensitive to peakedness and asymmetry of the distri-bution,respectively.Expression for optimal weighting be-tween the parts is derived using an efficacy measure.Theperformance of the proposed contrast and employed effi-cacy measure are studied in simulations.1.INTRODUCTIONThis paper deals with the problem how skewness of sourcedistributions can be exploited in Blind Source Separation(BSS)or Independent Component Analysis(ICA).Tradi-tionally,BSS methods assume implicitly that source dis-tributions are symmetric[1,2,3].However,in many ap-plication areas,such as in biomedical signal processingand telecommunications,the source distributions may beskewed.As an example,fast and slow fading encounteredin mobile digital communication systems are often charac-terized using Rayleigh and log-normal distributions that areboth asymmetric.In this paper we show how skewness information maybe used to improve the estimator needed infinding the inde-pendent components and consequently improve the qualityof separation.More dramatically,we demonstrate that insome cases ignoring the skewness information may lead tototal failure in separation.We propose an estimating func-tion based on absolute moments.It is composed of twoparts:one associated with lack of symmetry(skewness)and the other characterizing peakedness(kurtosis)propertywhere is the element of matrix .An ICA method consists of three parts:a theoretical measure of independence;an estimator(contrast,objective function)for the chosen theoretical measure,andfinally an algorithm for minimizing or maximizing the selected objec-tive function(see,[4]).Because we are interested in separating source distri-butions that may also be asymmetric,we consider contrast functions that can be presented as a sum of absolute val-ues of symmetric(even function)and asymmetric contrast (odd function).In general,these contrast functions can be presented as follows(3) where are weighting parameters between the sym-metric and the asymmetric contrast.Optimal values of the parameter are data dependent.In this section we pro-pose a choice for the contrast function and a method to de-termine the optimal’s iteratively.2.1.Optimal weighting in combined ICA contrastThe performance analysis of contrast functions is consid-ered in[4],[1],[5]and[6].It is usually assumed in the analysis that all sources are identically distributed.Local stability is found to depend on the following non-linear mo-ments(4) and the variance of separation solution is found to depend on(5) In[5]it is proposed that the following measure can be used as a performance criterionThe kurtosis of a distribution with unit variance can be measured by the third absolute moment(16) As a measure for skewness we can use the second skewed absolute moment(17) Exploiting and we may construct an ICA contrast. First,wefind that for Gaussian random variable with and(18) and.Furthermore,we define measures resem-bling the cumulant based kurtosis and skewness(20) The behavior of and appears to be analogous to the behavior of cumulant based kurtosis and skewness,ually,the sign of equals to the sign of and the sign of equals to the sign of.The polynomial or-ders of and are lower than the polynomial orders of and.This suggests that estimators of and have lower variance than estimators of and and smaller sample sizes are needed.We propose the contrast(21) The estimating function related to contrast(21)and the derivative of the estimating function are given bysign sign(22)sign sign sign(23)Based on the efficacy measure(6)the optimal weighting parameter for the absolute moment based contrast function can be given as followssign signsign signsign sign(24)sign signsign signsign sign(25) where is the th central moment,is the th absolute moment,is the th skewed absolute moment andsign sign(26)sign(27) Instead of trying to estimate the optimal weighting for sources,we estimate the optimal weighting parameters for the current data.Because is known,the statistics in (24)and(25)can be replaced by their sample counterparts. It is clear that this approach does not always lead to optimal estimates but in the weighting problem an approximate so-lution provides sufficiently good performance and the com-putation is straightforward.3.EXAMPLES3.1.Weighting between symmetric and asymmetric contrastsExamples on weighting between symmetric and asymmet-ric parts in the combined contrast are presented next.The converge of(24)and(25)is demonstrated in three differ-ent situations:Uniform distributed source(i.e.symmetric case)and two cases with different values of skewness.The convergence of the weighting parameters are presented as a function of the sample size in Figure1.3.2.Two source exampleQuality of separation of the proposed ICA criterion is stud-ied in simulation experiments.The absolute moments based contrast is straightforwardly implemented to FastICA al-gorithm[3].Other gradient type algorithms are suitable, parisons are made with the standard FastICA con-trasts.Differences in performance between symmetric con-trast and its asymmetric generalization are illustrated.As a05001000150020002500300035004000450050000.10.20.30.40.50.60.70.80.91number of observations|ω2| / (|ω1|+|ω2|)(a)Uniform distribution(b)GLD,,(c)GLD,,Fig.1.An example on convergence of the optimal weight-ing in the case of absolute moment contrast.Parameters (24)and (25)are estimated from a single source and the ratio is plotted.Uniform distribution is symmetric and thus the ratio of the weighting parameters converges towards zero when the number of observations increases.The other examples generated from Generalized Lambda Distribution (GLD)[8,9]are asymmetric and con-sequently the asymmetric estimating functions have large weights.The values presented are means over 100realiza-tions.benchmark we use a simple case where two skewed sources are mixed.The overall goal of the simulations is to demon-strate that the proposed contrast separates reliably both sym-metric and asymmetric sources.In the case of symmet-ric sources the the proposed asymmetric contrast reduces to the symmetric contrast and it is expected that its perfor-mance is approximately equal to that of conventional con-trasts.When the source distributions are asymmetric,it is expected that the asymmetric contrast will outperform the conventional contrasts.Both sub-Gaussian and super-Gaussian sources are used in simulations.The details of source signal statistics are given in Table 1and the den-sity functions of the theoretical distributions are presented in Figure 2.In each simulation the length of signals is 10000and the number of realizations is 1001.A full rank random mixing matrix is generated for each realization.SimulationSource 1Source 2pdfLaplace03BUniform 0-1.2GLD0.51DGLD0.3-0.5GLD-0.5-0.1FGLD0.30GLD0.8HGLD0.80.1Table 1.The theoretical third and fourth cumulants of source signals for the simulation experiments.The sources have zero mean and unit variance.In simulations from C to H,sources are generated from the Generalized Lambda Distribution (GLD)[8,9]with the corresponding theoret-ical cumulants.The density functions are visualized in in Figure 2.The quality of the separation is measured by Signal toInterference Ratio SIR(dB)MSE ,where MSE stands for Mean Square Error MSE ).To eliminate scaling differences both original signals and extracted signals are normalized to have zero mean and unit variance before the calculation of the SIR.After that source signals are matched to the extracted signals so that the re-sulting MSE values are as small as possible.Boxplots are used to describe the SIR-values in simula-tions.The boxplot is graphical presentation tool for sample distributions and it is widely utilized in applied statistics.In a boxplot the box defines the quintile range (from 25%percentile to 75%percentile).The line inside the box is the median.The ’whiskers’are lines extending from each end of the box to show the extent of the rest of the data.The length of a whisker is defined as 1.5times the length of the quintile range.Nevertheless,the whiskers are always bound by sample minimum and maximum.The possible outliers,outside the whiskers area,are marked by crosses.In Figure 3,the SIR values of the first extracted signals are presented.In simulations A-E all contrast functions didFig.2.The theoretical density functions of source distribu-tions.The letter of the simulation and the number of source are given on the left side of each plot.The cumulants of distributions are summarized in Table1.well:median SIR values are over30dB.In simulations F, G and H the symmetric contrasts are outperformed by the skewed absolute moment.The overall results indicate that the skewed absolute third moment contrast(3)are the most reliable among the methods considered.This is a reasonable result because even if the weighting parameter estimators (24)and(25)are not simple,they use only statistics up to the fourth order.These simulations are only simple special cases but they do,nonetheless,strongly support the idea that skewness information may significantly improve the quality of separation.4.CONCLUSIONWe considered blind separation using absolute moments as estimating functions.Absolute moments are computation-ally simple and have lower sample variance compared to cu-mulants.The proposed estimating function combines abso-lute moments(symmetric estimating function)and skewed absolute moments(asymmetric estimating function).Con-sequently,the separation remains good even if the source distributions are skewed.The optimal weighting between the symmetric and asymmetric estimating function is ob-using the concept of BSS efficacy.In practice,the is adaptively estimated from data.Simulationdemonstrate the reliable performance of the pro-method for various different sources.5.REFERENCESJean-Francois Cardoso and Beate Hvam Laheld,“Equivariant adaptive source separation,”IEEE Trans-actions on Signal Processing,vol.44,no.12,pp.3017–3030,Dec.1996.Shun-Ichi Amari,Andrzej Cichocki,and H.Yang,“A new learning algorithm for blind signal separation,”in Advances in Neural Information Processing Systems, vol.8,pp.757–763.MIT Press,Cambridge MA,1996.Aapo Hyv¨a rinen,“Fast and robustfixed-point algo-rithms for independent component analysis,”IEEE Transactions on Neural Networks,vol.10,no.3,pp.626–634,1999.Jean-Francois Cardoso,“Blind signal separation:Sta-tistical principles,”Proceedings of the IEEE,vol.86, no.10,pp.2009–2025,1998.[5]Saleem A.Kassam,Yinglu Zhang,and George V.Moustakides,“Some results on a BSS algorithm un-der non-standard conditions,”in Proc.of the33rd An-nual Conference on Information Sciences and Systems, 1999.[6]Jean-Francois Cardoso,“On the performance of orthog-onal source separation algorithms,”in Proc.EUPISCO, 1994,pp.776–779.[7]Alan Stuart and J.Keith Ord,Kendall’s Advanced The-ory of Statistics:Distribution Theory,vol.1,Edward Arnold,sixth edition,1994.[8]Zaven A.Karian,Edward J.Dudewicz,and Patrick Mc-Donald,“The extended generalized lambda distribution system forfitting distributions to data:History,comple-tion of theory,tables,applications,the”final word”on momentfits,”Communications in Statistics:Simulation and Computation,vol.25,no.3,pp.611–642,1996. [9]Jan Eriksson,Juha Karvanen,and Visa Koivunen,“Source distribution adaptive maximum likelihood esti-mation of ICA model.,”in ICA2000.Proceedings of the Second International Workshop on Independent Com-ponent Analysis and Blind Signal Separation,2000,pp.227–232.12320406080100S I RContrast(a)Simulation A(b)Simulation B(c)Simulation C (d)Simulation D(e)Simulation E(f)Simulation F(g)Simulation G (h)Simulation HFig.3.Boxplots of the SIR-values of first separated signal.The non-linearities in comparison are kurtosis (1),tanh (2)and skewed absolute moment (3).The source distributions used in simulations are summarized in Table 1and in Figure 2.。