[2006 TIM Acoustic imaging of underwater embedded]

合集下载

The effect of climate on acoustic signals中文翻译

气候变化对声学信号的影响：大气吸声是否会对鸟的歌声和蝙蝠的定位叫声造成影响？生态学和进化生物学系，图森，亚利桑那大学，1041东洛厄尔街，亚利桑那州85721 沿着生态梯度的信号差异可能会导致新物种生成。

目前的研究验证了这个假说，沿气候梯度的声音吸收的变化为声学信号的差异做出了选择，这不仅有助于理解物种的多样性，而且对了解生物如何应对气候变化也有帮助。

由于吸声会随温度、湿度和声音频率而变化，个体或物种的信号结构可能会随空间或时间上的气候变化而改变。

特别是具有较低频率、较窄带宽和较长持续时间的信号在高吸声环境中应该会更易检测到。

通过研究北美木莺和美国西南地区的蝙蝠，这项工作发现了信号结构和吸声之间存在关联的证据。

整个活动范围具有较高的平均吸收值的莺鸟更可能具有较窄带宽的歌声。

在具有较高吸收值的栖息地发现的蝙蝠种类更可能有较低频率的叫声。

此外，蝙蝠的定位叫声结构会随着季节而变化。

在具有较高吸声的雨季中，蝙蝠会使用具有较长持续时间、较低频率的叫声。

这些结果表明，尽管吸声的影响不算太大，但由于吸声的变化，信号可能会随着气候梯度发生改变。

1 绪论因为环境会影响声学和视觉信号从信号发送者到接收者的传输方式(Morton, 1975; Wiley 和Richards, 1978; Richards 和Wiley, 1980; Endler,2000)，所以信号系统往往会沿生态梯度发生改变(Hunter 和Krebs,1979; Wiley, 1991; Marchetti,1993;Badyaev和Leaf,1997;McNaught and Owens, 2002;Slabbekoorn and Smith, 2002; Tobias et al., 2010)。

这种改变对于物种的形成和多样化有重要意义，因为信号方面的变化可能导致合子前隔离(West-Eberhard, 1983; Endler, 1992;Schluter 和Price, 1993; Grant 和Grant, 1996; Irwin 等,2001)。

MMSE BEAMFORMER BASED ON PARTIAL FFT DEMODULATION FOR OFDM UNDERWATER ACOUSTIC COMMUNICATIONS

MMSE BEAMFORMER BASED ON PARTIAL FFT DEMODULATION FOR OFDM UNDERWATER ACOUSTIC COMMUNICATIONSGema Pi˜n eroInstitute of Telecommunicationsand Multimedia Applications Universitat Polit`e cnica de ValenciaValencia,Spain Email:gpinyero@iteam.upv.esAndrew C.SingerCoordinated Science Laboratory University of Illinois at Urbana-Champaign Urbana61801,IL,USAEmail:acsinger@ABSTRACTUnderwater Acoustic(UW A)communications have to deal with channels that exhibit widely time-spreading multipath responses, together with Doppler spreading and limited bandwidth.These features seriously limit data throughput,even for sophisticated modulation schemes.Focusing on the large time-spreading charac-teristics,we analyze in this paper an OFDM UW A system such that the channel is longer than the Cyclic Preﬁx(CP)used.Under this condition,and considering an array of hydrophones at the receiver, we propose in this paper a Minimum Mean Square Error(MMSE) beamformer that makes use of two partial FFTs of half the original number of subcarriers to demodulate the OFDM signal.We show that one of the two demodulated signal can be received without Inter-Symbol Interference(ISI),even for longer-than-CP channels, and that the MMSE beamformer automatically is steered to direct and reﬂected paths obtaining an improvement of4dB in the Signal-to-Noise Ratio(SNR)and a reduction of2to3times the Bit Error Rate(BER)with respect to a conventional Delay-and-Sum(DS) beamformer.1Index Terms—Underwater Acoustic Communications,broad-band acoustic beamformer,OFDM.I.INTRODUCTIONUnderwater Acoustic(UW A)communication channels present a multipath model together with particular features not common to other wireless communications[1],[2].They exhibit present time variations even when the transmitter and the receiver are anchored,due to the changing sea surface (waves),and the spatial variation of the propagation velocity which strongly depends on water temperature.Figure1 shows signal propagation where the direct path is summed up with reﬂected paths due to multiple surface and seabed reﬂections.Propagation attenuation increases with frequency, resulting in near5dB/Km loss for a carrier frequency of20kHz or12dB/Km at40kHz,which limits the useful bandwidth to a few kHz.Although small compared to other communications systems,the bandwidth of the information signal is usually broadband compared to the carrier frequency.1This work has been partially supported by Spanish Government through CYCIT grant TEC2009-13741and Prog.Nac.Movilidad RRHH2011,and by Regional Government through PROMETEO/2009/013.Fig.1.Signal propagation through a multipath underwater acoustic channel.Multipath in the UW A channel can be modeled as a time-varying channel impulse response h(t,τ).If the transmitter and receiver are stationary and the coherence time is large enough,i.e.h(t,τ) h(τ),h(τ)is a linear time-invariant (LTI)multipath channel.Let us consider the equivalent discrete-time model of the received signal y(t)when sam-pled with sample period T s=1/f s.The received discrete-time signal can be expressed through the convolution sum between the transmitted signal and the multipath channely(n)=L−1l=0h(l)x(n−l)where y(n)=y(nT s),x(n)= x(nT s),and L is the length,in number of symbols,of the discrete-time channel h(n).The discrete-time model of the channel is useful for the design of the equalizer and detector at the receiver.While similar in form to Radio-Frequency multipath channels,UW A channels are widely spread in time.The large delay spread together with small bandwidth severely limits the throughput in these communication sys-tems.A long-range UW A channel can exhibit a delay spread of200-400ms depending on the propagating environment [3].Regarding the receiver complexity,extremely long chan-nels require even longer equalizers,much more complex for the multi-channel case.One way to simplify the receiver is to use a beamformer in theﬁrst stage of the receiver,and to process the resulting signal through a Decision-Feedback Equalizer in a second stage[4],[5].Regarding OFDM communications in UW A channels, Cyclic Preﬁx(CP)length should be longer than channel20th European Signal Processing Conference (EUSIPCO 2012)Bucharest, Romania, August 27 - 31, 2012delay spread to avoid Inter-Symbol Interference(ISI),which in turn requires large OFDM symbols in order to not sacriﬁce rate.Throughout this paper,we have focused our work on LTI UW A channels considering any Inter-Carrier Interfer-ence(ICI)due to motion negligible.While the CP length in OFDM communications has to be larger than channel length,long-range communications may be impractical due to the channel delay spread.However,if a CP shorter than the channel is used,the circular convolution property of the Discrete-Time Fourier(DFT)is no longer respected and the received signal is no longer the product of the original symbol and the channel DFT.We will model the use of shorter CP as a random error added to each subcarrier,as we will show in Section II.Using this OFDM signal and error model,we will design a Minimum Mean Square Error(MMSE)beamformer that makes use of two DFTs of half the number of subcarriers. We will study the error model for this particular design when channel time-spreading is larger than the CP.Finally,we will compare the use of the MMSE beamformer with the conventional Delay-and-Sum(DS)beamformer[6]by means of numerical simulations.II.PARTIAL FFT DEMODULATION OF THEOFDM RECEIVED SIGNALLet us consider a single channel UW A communica-tion system transmitting linear modulated symbols through OFDM.Once the inverse DFT of the baseband modulated symbols X(k)is performed,a Cyclic Preﬁx(CP)formed by the last L CP samples of x(n)=DFT−1[X(k)]is added to form x CP(n).In absence of noise and interference,thereceived signal y CP(n)can be expressed as:y CP(n)=x CP(n)∗h(n)=L−1l=0h(l)x CP(n−l)(1)The last N samples of y CP(n)correspond to the cir-cular convolution between x(n)and h(n)denoted by x(n)N h(n).The DFT of the circular convolution of two signals is equal to the multiplication of their respective DFTs,which means that symbols X(k)are multiplied by the complex amplitude of the corresponding subcarriers:y(n)=h(n)N x(n)DFT←→Y(k)=H(k)X(k)(2) where Y(k)is the DFT of the last N samples of y CP(n), denoted by y(n).The use of a DFT that processes separately small sets of subcarriers to improve Doppler selectivity has been recently proposed in[7].It has been shown to be computationally inexpensive,and it presents good performance for DFTs of N/4and N/8points,N being the total number of subcarriers of the OFDM symbol.In our study,Doppler selectivity is considered to have been compensated in a previous stage(by resampling the signal as,for example, in[8]),and our study has focused on the use of DFTs of N/2points to improve the Signal-to-Noise Ratio(SNR)and Bit Error Rate(BER)in longer-than-CP channels.Let us deﬁne two sequences of N/2points x1(n)and x2(n)obtained from x(n)as:x1(n)=x(n),n=0,...,N/2−1(3)x2(n)=x(n+N/2),n=0,...,N/2−1(4)that is,x1(n)is formed by theﬁrst N/2points of OFDM sequence x(n)and x2(n)is formed by the last N/2points. Consider now the deﬁnition of two similar sequences over the received OFDM signal y(n)once the CP has been dropped,that is,y1(n)=y(n),n=0,...,N/2−1and y2(n)=y(n+N/2),n=0,...,N/2−1.Given the expressions of x1(n)and x2(n)in(3)-(4), signal x1(n)can be considered as an OFDM signal whose preﬁx is formed by the last L CP samples of x2(n),whereas signal x2(n)is considered to have a preﬁx formed by the last L CP samples of x1(n).One way to constrain x1(n)and x2(n)to fulﬁll the circu-lar convolution property is to use a particular pilot sequence X(k)such that its inverse DFT holds x1(n)=x2(n). This implies that X(k)=DFT[x(n)]must be null at odd values of k.Assuming that x1(n)=x2(n),equations(5)-(7)summarize the different relationships between sequences and their corresponding DFTs,considering in all sequences a CP larger or equal to the channel length(L CP L):y(n)=x(n)N h(n)DFT←→Y(k)=X(k)H N(k)(5) y1(n)=x1(n)N/2 h(n)DFT←→Y1(k)=X1(k)H N2(k)(6) y2(n)=x2(n)N/2 h(n)DFT←→Y2(k)=X2(k)H N2(k)(7) where X1(k)=DFT[x1(n)],X2(k)=DFT[x2(n)], H N(k)is the DFT of N points of h(n),and H N2(k)is the corresponding DFT of N/2points of h(n).Both DFTs are related through H N2(k)=H N(2k),k=0,1,...N/2−1, as long as L<N/2.It can be stated that y1(n)and y2(n) are the respective circular convolution of x1(n)and x2(n) with channel h(n).In the case of channels spreading further than the CP but less than N/2,that is,L CP<L and L<N/2,the last N samples of the received sequence y CP(n)given by (1)no longer represent the circular convolution between x(n)and h(n):theﬁrst L−1−L CP samples of y(n) are different from the corresponding ones of h(n)N x(n), whereas the last N−(L−1−L CP)are equal.The same effect can be observed on y1(n).However y2(n)does fulﬁll the property in(7)because the cyclic preﬁx of x2(n)is formed by sequence x1(n)and extends over N/2.Therefore, y2(n)can be expressed as in(7),whereas y1(n)is givenE . . . S/P (n)DFTN/2 pointsS/P (0,…,N/2-1). . .DFTN/2points(n)w (k)+x-Y Y DropCP and split(n)Fig.2.Block diagram of the MMSE beamforming.by y 1(n )=x 1(n )Nh (n )+q (n )where q (n )is an error sequence of length L −1−L CP .In fact,q (n )is the Inter-Symbol Interference (ISI)produced by the multipath channel that is not cancelled because of L >L CP .The corresponding DFT of y 1(n )can then be expressed as:Y 1(k )=X 1(k )H N 2(k )+Q (k )(8)where Q (k )is the DFT of N/2points of the sequence q (n ).It has to be noted that the ISI error q (n )is conﬁned to the ﬁrst L −1−L CP samples in time,whereas Q (k )is spread over all frequencies from k =0to k =N/2−1.In the next section we will make use of the different expressions of Y 1(k )and Y 2(k )to design an MMSE beamformer at the receiver avoiding the requirement of knowing the angles of arrival of the multiple paths.III.MINIMUM MEAN SQUARED ERROR (MMSE)BEAMFORMER BASED ON PARTIAL FFT In Section II,the model of a single OFDM-UW A channel has been stated.We introduce now a multi-channel commu-nication system that makes use of multiple hydrophones at the receiver.From Figure 1,it can be assumed that different paths arrive to the hydrophone array from different angles,called Directions-of-Arrival (DoA),and are denoted by θ.The DoA is deﬁned as the angle of arrival with respect to the line of the hydrophones for a linear array.Let us assume that a coarse estimate of the DoA of the direct path in Figure 1is calculated (for example,by means of 2-D frequency-θspectral estimation of the received data),and denote that direction as θd .The block diagram of the proposed scheme is depicted in Figure 2where a (θd ,k )is the steering vector [6]for DoA θd and frequency k ,and a linear array of M hydrophones has been considered.The M components of the steering vector a (θd ,k )depend on the array geometry,the nominal frequency,the propagation velocity,and on θd .Signal y mCPis the signal recorded at the m -th hydrophone such that (1):y mCP=x CP ∗h m (n )+v m (n )(9)where v m (n )is an additive complex white Gaussian noiseof zero mean and variance σ2n,and h m (n )is the channel impulse response between the transmitter and the m -th hydrophone.As it can be seen from Figure 2,signals y mCPare reduced to their last N samples by dropping the Cyclic Preﬁx.They are subsequently split in two to compute two partial DFTs of N/2points obtaining sequences Y m 1(k )and Y m 2(k ),for k =1,...,N/2.From this point on,hydrophone outputs are collected together at each k in order to form the following spatial vectors:Y i (k )= Y 1i (k ),...,Y m i (k ),...,Y M i (k ) Ti =1,2(10)where superscript (·)T stands for transposed.Then,a H (θd ,k )Y 2(k )is used as the reference signal,whereas Y 1(k )is ﬁltered by the narrowband beamformer,w (k )=[w 1(k ),...,w m (k ),...,w M (k )]T .The MMSE beamformer of Figure 2for frequency k can be expressed as :w MMSE (k )=min wE [|E (k )|2](11)where E (k )is the error signal formed as:E (k )=a H (θd ,k )Y 2(k )−w H (k )Y 1(k )(12)where superscript (·)H stands for transposed and conjugated.Substituting equation (12)into (11),the MMSE weights can be calculated as:w MMSE (k )=R −1Y 1(k )R Y 1Y 2(k )a (θd ,k )(13)where R Y 1(k )is the autocorrelation matrix of vector Y 1(k )and R Y 1Y 2(k )is the correlation matrix between vectors Y 1(k )and Y 2(k ).Both matrices can be estimated from a number N T M of respective snapshots assuming stationary signals within a period N T ·N ·T s seconds.The general expression of the estimated correlation matrix for column vectors Y i (k ),Y j (k )is given by:R Y i Y j (k )=1N T N T n =1Y (n )i (k )Y (n )j (k )H (14)where Y (n )(k )denotes the spatial vector formed by k -th DFT coefﬁcients calculated from the OFDM symbolpresented at the hydrophones at time n .Considering now the models of Y 1(k )and Y 2(k )given in (8)and (7)respectively,and assuming that pilot signals X 1(k )=X 2(k )have unit variance and are uncorrelated with all the noise signals V m 1(k )and V m 2(k ),and with error Q (k ),the MMSE beamformer (13)is formed by the following terms:w MMSE=(R H+R V1+R Q)−1R H a(θd)(15) where R H=E[H(k)H H(k)]is the autocorrelation ma-trix of the channel spatial vector,whose components are the channel values at each hydrophone,H(k)=[H1(k),...,H M(k)]T.R V1is the autocorrelation matrix ofnoise spatial vector V1(k)=[V11(k),...,V M1(k)]T,and R Q is the autocorrelation matrix of spatial error vector due to ISI,Q(k).The MMSE beamformer minimizes the difference between the reference signal,a H(θd,k)Y2(k)and the beamformer output.As the main difference between both signals is due to the noise and the ISI Q(k),the resulting beamformer will try to reduce both contributions.To better understand its behavior,let us assume a simple UWA channel with only direct and reﬂected paths,denoted by h d(n)and h r(n)respectively.Thus,spatial channel vector can be expressed as H=a(θd)H d+a(θr)H r,where the dependence on k has been dropped for the sake of clarity.As H d and H r can be assumed uncorrelated,the autocorrelation matrix of H can be expressed as:R H=E[|H d|2]a(θd)a H(θd)+E[|H r|2]a(θr)a H(θr)(16) Substituting(16)into the last term of(15)we obtain:R H a(θd)=M E[|H d|2]a(θd)+αM E[|H r|2]a(θr),(17) where a H(θd)a(θd)=M andα=a H(θr)a(θd)/M is a complex scalar such that|α| 1.Substituting(17)in(15), the MMSE beamformer can be expressed as:w MMSE=M(R H+R V1+R Q)−1[E[|H d|2]a(θd)+ +αE[|H r|2]a(θr)](18) The statistical analysis of MMSE beamformer is out of the scope of this paper,but from expression(18)it can be seen that MMSE beamformer has a direct dependence on both direct and reﬂected paths,avoiding a wrong performance when dealing with multipath(correlated)signals.IV.NUMERICAL SIMULATIONSIn this section the performance of the proposed MMSE beamformer for a multichannel UW A OFDM communi-cation system that makes use of N=512subcarriers is evaluated.The channel is composed of three different diffuse paths(similar to the plot of Figure1),a direct signal arriving fromθd=90◦and spreading fromτ=0·T sym to τ=7·T sym,aﬁrst reﬂected signal arriving fromθr1=115◦and spreading in time fromτ=6·T sym toτ=L·T sym, and a second reﬂected signal arriving fromθr2=45◦and spreading in time fromτ=16·T sym toτ=L·T sym, where L is the total length of the channel in symbols and T sym is the PSK symbol period.Channel length varies in theChannel LengthOutput SNR − Input SNR = 25 dBFig.3.SNR for theﬁrst hydrophone(without beamforming), and after MMSE and DS beamforming.Units are in dB.range L⊆(32,54)·T sym.Notice that both reﬂected paths are simulated to be of length L,since,in a realistic environment, usually both seabed and wave reﬂections cause symbol time-spreading.Direct and reﬂected paths have individual taps h(τ)modeled as independent complex Gaussian of zero mean and unit variance atτ=l·T sym.A source of BPSK pilot symbols X(k)such that x1(n)= x2(n)is transmitted using OFDM with aﬁxed Cyclic Preﬁx length of32symbols.The pilot sequence is used as a probe to initially steer the beamformers.A linear array of M=10hydrophones with half wavelength spacing has been used and white Gaussian noise has been added at each hydrophone resulting in a received SNR of25dB. Finally,MMSE beamformer of Section III and Delay-and-Sum(DS)beamformer given by w DS(k)=a H(θd,k)have been calculated for300independent runs.Figure3shows the output SNR for theﬁrst hydrophone (without beamformer)and after MMSE and DS beamform-ing respectively.It can be seen that the SNR obtained by the MMSE is3-4dB higher than DS and8dB higher than input SNR(ﬁrst sensor)for the whole range of channel lengths L. This behavior agrees with the Array Gain(AG)expected for an array of10sensors separated d=λ/2[3],once the SNR has been maximized.Another interesting feature is the capacity of the beam-former to reduce the contribution of the channel components outside the CP.Figure4shows the energy ratio between channel taps allocated outside the CP and those placed within the CP forﬁltered MMSE and DS channels and for the chan-nel seen at theﬁrst hydrophone.DS decreases this ratio in2 dB and MMSE only in1dB,but considering that reﬂected paths produce all the outside-CP energy,and that MMSE has a signiﬁcant increase in SNR due to their contributions,it can be stated that MMSE is actually decreasing the energy placed outside CP with respect to the original channel. Finally,the Bit Error Rate(BER)has been calculated for100independent runs.For each run,2.5·104uncoded−−−−−−−Channel LengthOut −to −within −CP Energy ratioFig.4.Energy ratio between channel taps outside the CP and those placed within the CP:original (ﬁrst hydrophone)and after MMSE and DS beamforming.Units are in dB.bits have been mapped into QPSK symbols and have been transmitted by means of 26OFDM symbols.For all symbols,20subcarriers at each side have not been used forming a guard band of null content.Previously,a probe formed by 12pilot OFDM symbols has been used to design the MMSE beamformer.From the probe,estimated channel coefﬁcients ˆH(k )have been calculated for MMSE and DS ﬁltered signals,and for the signal received at the ﬁrst hydrophone as well.As the pilot sequence is such that X (k )=0for oddfrequencies,estimated coefﬁcients ˆH(k )have been directly obtained only for even k .Afterwards,coefﬁcients ˆH(k )for odd k have been calculated through linear interpolation.Figure 5shows the BER obtained for two different channel lengths,L =35and L =50.Both ﬁgures show a reduction of the BER when using MMSE instead of DS beamforming,achieving a signiﬁcant decrease of 3-4times for high SNRs.V .CONCLUSIONAn MMSE beamformer based on two partial DFTs of N/2subcarriers has been proposed for improving the received SNR and BER when dealing with widely time-spreading UWA channels.A simple analysis for one direct and one reﬂected path have shown that MMSE beamformer is auto-matically steered to both DoAs,whereas conventional DS only enhances the direct path.Numerical simulations have conﬁrmed this beneﬁt obtaining BER reductions of 2to 4times depending on the received SNR and the channel length.VI.REFERENCES[1]M.Stojanovic and J.Preisig,“Underwater acoustic communi-cation channels:Propagation models and statistical characteri-zation,”Communications Magazine,IEEE ,vol.47,no.1,pp.84–89,January 2009.[2] A.Singer,J.Nelson,and S.Kozat,“Signal processing forunderwater acoustic communications,”Communications Mag-azine,IEEE ,vol.47,no.1,pp.90–96,January 2009.[3]T.Yang,“A study of spatial processing gain in underwateracoustic communications,”Oceanic Engineering,IEEE Journal of ,vol.32,no.3,pp.689–709,july 2007.101010EbN0 (dB)BER(a)Channel length,L =35101010EbN0 (dB)BER(b)Channel length,L =50Fig.5.BER obtained for OFDM modulated QPSK symbols and different channel lengths.Equal CP length of L CP =32symbols has been used for both channels.[4]M.Stojanovic,J.Catipovic,and J.Proakis,“Reduced-complexity simultaneous beamforming and equalization for underwater acoustic communications,”in OCEANS ’93.En-gineering in Harmony with Ocean.Proceedings ,oct 1993,pp.III426–III431vol.3.[5]P.-P.Beaujean and L.LeBlanc,“Adaptive array processing forhigh-speed acoustic communication in shallow water,”Oceanic Engineering,IEEE Journal of ,vol.29,no.3,pp.807–823,july 2004.[6] D.H.Johnson and D.E.Dudgeon,Array signal processing:Concepts and techniques .Prentice-Hall,Signal Processing Series,1993.[7]S.Yerramalli,M.Stojanovic,and U.Mitra,“Partial fft demodu-lation:A detection method for doppler distorted ofdm systems,”in Signal Processing Advances in Wireless Communications (SPAWC),2010IEEE Eleventh International Workshop on ,June 2010,pp.1–5.[8]L.Baosheng,Z.Shengli,M.Stojanovic,L.Freitag,and P.Wil-lett,“Non-uniform doppler compensation for zero-padded ofdm over fast-varying underwater acoustic channels,”in OCEANS 2007-Europe ,June 2007,pp.1–6.。

一种新型极低比特率声码器在音素HMM语音识别中的应用

一种新型极低比特率声码器在音素HMM语音识别中的应用李颖;张有为
【期刊名称】《五邑大学学报（自然科学版）》
【年(卷),期】1999(013)004
【摘要】音素ＨＭＭ语音识别是当前语音识别领域的一个热点。

本文在简单介绍音素识别的相关理论后，着重讨论了一种新的极低速率语音编解码器的原理，方法和在音素识别上的应用，最后给出相关的实验结果。

【总页数】5页(P37-41)
【作者】李颖;张有为
【作者单位】五邑大学为电子与信息工程系;五邑大学为电子与信息工程系
【正文语种】中文
【中图分类】TN912.34
【相关文献】
1.一种CDHMM/MLP混合模型及其在语音识别中的应用 [J], 史笑兴;顾明亮;王太君
2.一种高质量的极低比特率声码器 [J], 张志伟;吴家安
3.一种高质量的极低比特率声码器 [J], 张志伟;吴家安
4.矢量量化法用于改进的多带激励声码器一种极低比特率语音编码方案 [J], 刘波涛;匡镜明
5.一种新型的滤波器组在水声应答器中的应用 [J], 彭会斌
因版权原因，仅展示原文概要，查看原文内容请购买。

数字语音信号处理教案

数字语音信号处理实验指导书前言语音信号处理是研究用数字信号处理技术和语音学知识对语音信号进行处理的新兴的学科，是目前发展最为迅速的信息科学研究领域的核心技术之一。

通过语音传递信息是人类最重要、最有效、最常用和最方便的交换信息形式。

同时，语言也是人与机器之间进行通信的重要工具，它是一种理想的人机通信方式，因而可为信息处理系统建立良好的人机交互环境，进一步推动计算机和其他智能机器的应用，提高社会的信息化程度。

语音信号处理是一门新兴的学科，同时又是综合性的多学科领域和涉及面很广的交叉学科。

虽然从事这一领域研究的人员主要来自信号与信息处理及计算机应用等学科，但是它与语音学、语言学、声学、认知科学、生理学、心理学等许多学科也有非常密切的联系。

20世纪60年代中期形成的一系列数字信号处理的理论和算法，如数字滤波器、快速傅立叶变换（FFT）等是语音信号数字处理的理论和技术基础。

随着信息科学技术的飞速发展，语音信号处理取得了重大的进展：进入70年代之后，提出了用于语音信号的信息压缩和特征提取的线性预测技术（LPC），并已成为语音信号处理最强有力的工具，广泛应用于语音信号的分析、合成及各个应用领域，以及用于输入语音与参考样本之间时间匹配的动态规划方法；80年代初一种新的基于聚类分析的高效数据压缩技术—矢量量化（VQ）应用于语音信号处理中；而用隐马尔可夫模型（HMM）描述语音信号过程的产生是80年代语音信号处理技术的重大发展，目前HMM已构成了现代语音识别研究的重要基石。

近年来人工神经网络(ANN)的研究取得了迅速发展，语音信号处理的各项课题是促进其发展的重要动力之一，同时，它的许多成果也体现在有关语音信号处理的各项技术之中。

为了深入理解语音信号数字处理的基础理论、算法原理、研究方法和难点，根据数字语音信号处理教学大纲，结合课程建设的需求，我们编写了本实验参考书。

本本参考书针对教学大纲规定的四个研究设计型实验，每个实验给出了参考程序，目的是起一个抛砖引玉的作用，学生在学习过程中，可以针对某一个实验进行延伸的创新学习，比如说，语音端点的检测、语音共振峰提取、基于HMM或DTW的有限词汇或大词汇的特定人、非特定人的语音识别、识别率的提高（如何提高有噪环境下的识别率）、以及编码问题等，同时在学习中还可深入思考如何将有关的方法在嵌入式系统或DSP 下的实现问题等。

用微波在水表面的反射来检测水下声源

用微波在水表面的反射来检测水下声源
D.E.Tremain;D.J.Angelakos;李士才
【期刊名称】《声学技术》
【年(卷),期】1975(0)1
【摘要】我们做了一系列比较简单的实验,其中的一个实例如图1所示。

利用微波辐射波束进行发收分置情况下的散射测量来研究检测水下低频小功率声源的可能性。

通常,浸没于水下的声源为装在水平方板上的圆形膜片。

该膜片以45赫至55赫之间的频率振动,最大振幅可以是0.16厘米,也可以是0.08厘米。

用吹风机扰动水表面,产生一个骚动的水面,同时在水面上还叠加有由浸没于水中的声源所产生的小振
幅振动。

【总页数】5页(P175-179)
【关键词】自相关函数;实际位移;声源;噪声信号;检波器;仪表;周期分量;反向散射;微波波束;波动;滤波器系统;镜面反射;周期函数;超越函数
【作者】D.E.Tremain;D.J.Angelakos;李士才
【作者单位】加利福尼亚大学电子学研究所及电气工程和计算机科学系
【正文语种】中文
【中图分类】F2
【相关文献】
1.水下声源引起的水表面横向微波的理论研究 [J], 戴振宏;孙金祚;隋鹏飞
2.双谱幂次水下检测空中高速运动声源方法 [J], 韩建辉;杨日杰;王伟;林丽静
3.水下声场激励水表面横向微波的色散关系 [J], 宫彦军;毕冬梅;吴汉华;江荣熙;孙金祚;吴振森
4.模态信息非完备采样对水下声源检测的影响及改进方法 [J], 李明杨;孙超;邵炫
5.水下运动声源的方位方差检测方法 [J], 陈韶华;赵冬艳;郑伟
因版权原因，仅展示原文概要，查看原文内容请购买。

image alignment and stitching a tutorial

Image Alignment and Stitching: A Tutorial1
Richard Szeliski Last updated, December 10, 2006 Technical Report MSR-TR-2004-92
This tutorial reviews image alignment and image stitching algorithms. Image alignment algorithms can discover the correspondence relationships among images with varying degrees of overlap. They are ideally suited for applications such as video stabilization, summarization, and the creation of panoramic mosaics. Image stitching algorithms take the alignment estimates produced by such registration algorithms and blend the images in a seamless manner, taking care to deal with potential problems such as blurring or ghosting caused by parallax and scene movement as well as varying image exposures. This tutorial reviews the basic motion models underlying alignment and stitching algorithms, describes effective direct (pixel-based) and feature-based alignment algorithms, and describes blending algorithms used to produce seamless mosaics. It closes with a discussion of open research problems in the area.

多声道音频信号的时间与空间成形[发明专利]

专利名称：多声道音频信号的时间与空间成形
专利类型：发明专利
发明人：萨沙·迪施,于尔根·赫勒,马蒂亚斯·诺伊辛格,耶罗恩·布里巴特,杰拉德·霍特胡
申请号：CN200680037901.1
申请日：20060831
公开号：CN101356571A
公开日：
20090128
专利内容由知识产权出版社提供
摘要：由从具有高时间分辨率的采样值所组合而成的帧所表示的多声道信号的所选声道，当表示该所选声道的中间分辨率表示的波形的波形参数表示被推导出来时，该所选声道可以被编码成具有较高质量的形式；该波形参数表示包括具有比采样值的时间分辨率要低并且比帧重复率所定义的时间分辨率要高的时间分辨率的中间波形参数的序列。

具有中间分辨率的波形参数表示可以用于对重建声道进行成形，以获得具有与该所选原始声道的信号包络十分接近的信号包络的声道。

该成形过程是在比逐帧处理的时间刻度要短的时间刻度上执行，因此可以提高该重建声道的质量。

另一方面，该成形时间刻度比采样值的时间刻度要大，因此可以显著地降低波形参数表示所需的数据总量。

申请人：弗劳恩霍夫应用研究促进协会,皇家飞利浦电子股份有限公司
地址：德国慕尼黑
国籍：DE
代理机构：中科专利商标代理有限责任公司
代理人：朱进桂
更多信息请下载全文后查看。

686 RAPID COMMUNICATION nnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnn n

important additional findings like common iliac vein or inferior vena cava thrombosis (Fig.4).In two further cases,unknown contralateral thrombosis was seen with True-FISP MR-venog-Fig.1True-FISP MR-venography (TR 5.0ms,TE 2.5ms,860 860 m in-plane resolution)in a 42year old patient without thrombosis.Veins and arteries are displayed with a bright signal.A black rim surrounding the vessel lumen is readily apparent (magnification).True-FISP sequence for MR-venography,which may allow for a high-contrast MR-venography without costly contrast agent.Technical considerations of MR-venography using high resolution True-FISPTrue-FISP sequences demonstrate a T1/T2contrast due to data acquisition in a steady state condition of the transverse magnetization [9].Therefore,the long T2of blood allows for Fig.2True-FISP MR-venography (TR 5.0ms,TE 2.5ms,860 860 m in-plane resolution)in a 31year old pregnant plete left femoral (a )and iliac (b )vein thrombosis is diagnosed because of the dark signal,whereas open superficial and deep right veins show a bright signal.The inferior vena cava is highly compressed due to pregnancy (arrow in c ).in the inferior vena cava(Fig.4),a highly important finding for therapy.In two other patients,tumor masses causing venous compression were diagnosed on the MR-images.This shows the Recently,new MR-approaches for fast detection ofwith fast MR-venography。

所示。即基底膜的不同部位只对声波...

基于听觉外周模型与高阶统计量的语音流检测哈尔滨工程大学硕士学位论文基于听觉外周模型与高阶统计量的语音流检测姓名:申丽然申请学位级别:硕士专业:计算机应用技术指导教师:李雪耀20030201哈尔滨.科人学硕十学位论文摘要语音流检测是语音信号处理的重要内容之~,它直接关系到语音识别系统的性能。

至今为止,传统的语音流检测技术在安静环境下或较为稳定的噪声环境下,具有优良的效果,但在随机多变的恶劣环境下,检测的效果不尽理想。

本文研究的内容是各种随机噪声下的语音流检测。

所用的实验数据来自真实现场的录音。

涉及噪声种类繁多,如脉冲噪声、周期噪声、高斯噪声白、有色、非高斯噪声及其更为复杂的非平稳噪声。

本文主要工作如下:对人耳的听觉特性及人的听觉心理用于噪声抑制作了较为深入的研究。

用滤波器去掉相对于语音信号的频谱不变或缓慢变化的通道噪声,在此基础上用.技术对语音流进行了特征提取。

利用高阶累积量对高斯噪声的免疫能力,特别是对对称分布噪声的压制能力,研究了高阶累积量在语音流检测中的应用。

针对非平稳、非高斯的情况,本文提出了一种新的语音特征提取方法一一方差谱法。

它充分利用了语音信号的短时平稳特性和短时自相关特性。

能够有效的将语音信号和噪声信号分丌。

对以上三种特征提取的方法作了有机的结合,使他们优势互补达到更佳的效果。

把具有很好的泛化能力的支持向量机用于语音流和噪声的分类。

基于真实数据的大量的实验表明本文提出的算法很好的抑制了各种噪声优于传统的语音流检测方法。

关键词:语音流检测;高阶累积量;?;听觉感知;支持向量机哈尔滨:程人学硕十学位论文. ...,..,,, .,:/ ?,。

.,,? ,..一 ,, .;;?;:哈尔滨‘释大学硕学位论文第章绪论.课题研究的目的和意义语音流检测是一个古老的问题,最早可以追溯到酌世纪,但至今没有得到完全的解决。

语音流检测在语音信号处理中有着举足轻重的地位。

只有在『确的检测出语音流的情况下,其它的后续工作诸如语音识别、话者别爿有意义。

基于虚拟仪器的水声信号采集与处理

第28卷第4期增刊2007年4月仪器仪表学报Chinese Journal of Scientific Instr umentVol128No14Apr12007基于虚拟仪器的水声信号采集与处理张自嘉,王昌明,刘　伟(南京理工大学机械学院精密仪器系　南京　210094)摘　要:利用数据采集卡和La bVI EW软件对水声信号进行采集和处理,实现了对水下特定目标的声频特性分析。

利用由8个换能器组成的水听器阵,通过8通道数据采集卡对水下目标的噪声进行采集,实现了对水下目标定位的初步研究和算法验证。

利用虚拟仪器方便地实现了对水声信号的采集、处理,以及对水下目标测量时,为传感器结构和算法的确定提供参考。

关键词:虚拟仪器;水声信号;数据采集;波束形成Data acquisition and analysis of the under w a ter acoust ic signa lba sed on vir tual instr umentZhang Zijia,Wa ng Changmi ng,Li u Wei(De par tment of I n str ument of N a njing U nive rsit y of Science a nd Technology,N a njing210094,Chi na)Abstract:Underwater acousti c signal i s acquired and analyze d by using dat a aquisit ion cards a nd LabV IEW, and t he frequency charact erist ic of some underwat er sound source is i nvest igat ed.The underwater localization and correspondi ng arit hmet ic i s st udied t hrough82cha nnel s dat a acqui sition of t he noi se of t he underwater ob2 ject by use of eight hydrophone s.It i s convenient t o use t he virt ual i nst rument for dat a acqui sition and a naly2 si s of t he underwater acoustic signal,and help us to det ermine t he st ruct ure and t he arit hmetic for underwater measurement and det ect ion.K ey w or ds:vi rt ual i nst r ument;underwat er acoustic si gnal;dat a acquisit io n;bea mforming1　引言在雷达目标探测和水下声信号的采集、测量及水下目标探测等领域,通常需要多个换能器,并且需要对信号进行前置放大、数模转换、存贮和数字信号处理等,而且数据量较大。

基于回音壁模式光学微腔的细胞内传感

基于回音壁模式光学微腔的细胞内传感基于回音壁模式光学微腔的细胞内传感摘要：光学微腔作为一种新型的生物传感器，在生命科学中得到了广泛的应用，其基于回音壁模式的性质使得它可以实现高灵敏、高效率、非侵入式的细胞内传感。

本文将以回音壁模式光学微腔为核心，介绍其在细胞内传感方面的应用研究进展和主要应用场景。

首先，我们将介绍光学微腔回音壁模式的原理和特点，以及其与细胞生物学的关系。

接着，我们将详细讨论基于回音壁模式光学微腔的细胞内传感技术，包括气体、温度、压力、形态、质量、折射率等多项细胞内参数的测量原理和方法。

最后，我们将对该领域的研究现状和未来发展方向进行分析和展望。

关键词：光学微腔；回音壁模式；细胞内传感；气体；温度；压力；形态；质量；折射率引言：随着生命科学研究的深入和发展，对细胞内过程和环境的研究需求越来越迫切。

传统的分析方法需要对细胞进行破坏性取样分析，使得无法实时、动态地监测细胞内环境变化的情况。

因此，如何实现高灵敏、高效率、非侵入式的细胞内传感成为了一个重要的研究课题。

光学微腔作为一种新型的生物传感器，可以通过与物质相互作用而改变光场，从而实现对物质进行测量和传感。

其中，基于回音壁模式的光学微腔由于具有高灵敏度、高分辨率、免标记、实时监测等优点，被广泛应用于细胞内传感领域。

回音壁模式光学微腔的原理和特点光学微腔是一种光学共振器，在其内部，光波与微腔壁面多次反射形成驻波，即回音壁模式。

光子在光学微腔壁上来回反射的过程中，会受到微腔壁材料的光学性质影响，使得光子的相位和振幅发生变化。

这种光信号的微弱变化可以通过光纤耦合器和高灵敏光探测器测量，从而实现对光学微腔内环境参数的测量。

其中，回音壁模式是光学微腔的核心特点之一。

回音壁模式由于存在大量的回波，可以将光信号已在频域表示，从而实现高分辨率和高灵敏度的高精度测量。

此外，光学微腔还具有免标记、实时监测、高可重复性等优点，使得它成为了一种非常实用的细胞内传感器。

声学成像算法

声学成像算法
摘要：
一、声学成像算法的定义与原理
1.声学成像的定义
2.声学成像算法的原理
二、声学成像算法的发展历程
1.传统声学成像技术
2.现代声学成像算法的发展
三、声学成像算法的应用领域
1.医学成像
2.工业检测
3.地质勘探
4.其他应用
四、声学成像算法的技术挑战与发展趋势
1.技术挑战
2.发展趋势
正文：
声学成像算法是一种基于声波在物体内部传播的原理，通过对声波的接收和处理来实现对物体内部结构的无损检测和成像的技术。

声学成像算法的原理主要基于波动方程和声学逆问题，通过求解声波在物体内部传播的方程，以及对接收到的声波信号进行处理，从而实现对物体内部结构的成像。

声学成像技术的发展经历了从传统的基于射线追踪的方法到现代的基于数值模拟和机器学习的方法的转变。

近年来，随着计算机技术的快速发展，声学成像算法在医学成像、工业检测、地质勘探等领域得到了广泛应用。

在医学成像领域，声学成像算法可以用于实现对心脏、血管等内部结构的成像，为临床诊断提供重要依据。

在工业检测领域，声学成像算法可以用于检测材料内部的缺陷和裂纹，提高产品质量。

在地质勘探领域，声学成像算法可以用于探测地下资源，提高勘探效果。

尽管声学成像算法在实际应用中取得了显著成果，但仍然面临一些技术挑战，如信号处理、噪声抑制、成像分辨率等。

一种利用非正弦信号的水下声成像方法

ｎｌｎｅｌａｏｅｔｎｍｉｅｉｎｌｓｏｔｉｅ．Ｔｅｅｖｌｐｆｈｏｅａｅｉａｅｘｒｃｅｏｆｒａｄａｒｐｉｆｈｒｓｔｄｓｇａｉｂａｎｄｈｎｅｏｅｏｅｃｒｌｔｄｓａｃｔａｔｔｎｇｌｓｔｎｅｔｔｄｔｍｉｈａｏ
Ａｂｓｒｃ：Ｍｕｔ— ｅｍｍａｉｇｓｎｒｍｕｔａｏｄｅｅｎｔｎｔｒａｅｎｏｉｏｅＯｔａｒｔｎｏｅａｓｔａｔｌｉｂａｉｇｎｏａｓｖｉｌｍｅｓｉｈｅａｒｙｂｉｇｐｓｔｎｄＳｈｔｇａｉｇｌｂｓｃｕｅｉｅｆｃｓｏｍａｅ．Ｇｒｔｎｏｓａｅａｅｓｏｈｎｅｎｉｏｅ，ｏｈｉｅｍｆｒｃｐｉｎ，ｙｔｓｏｎｆｅｔｎｉｇｓａｉｇｌｂｅｒｒａｆｔｅａｔｎａｇｓｄｅｌｂｆｔｅｍａｎｂａｏｅｅｔｏｅｈｗｉｇｐａｓｉｏｒｄｎｉ．Ｗｈｎａｔｎａｅｅｎｔｓａｉｇｉｃｒａｅｈｎａｈｌｖｌｎｔｅｋｎｐｗｅｅｓｔｙｅｎｅｎｌｍｅｐｃｎｓｍｕｈｇｅｔｒｔａａｆｗａｅｅｇｈ，ｔｅａｉｓｎｆｅｔｈｌｉｇｅｆｃａｃｕｅｏａｉｇｌｂｓｔｐｏｃｈｅｅｆｔｉｏ．Ｔｏｓｌｅｔｉｒｂｅ，ａｎｒｔｒａｏｓｉｍ— ａｓｓｓｍｅｇｔｎｏｅｏａｐｒａｈｔｅｌｖｌｏｍａｎｌｂｅｒｈｅｏｖｈｓｐｏｌｍｎｕｄｅｗａｅｃｕｔｉｃａｉｔｏｔｏｓｎｓｉｌｓｇａｓｗａｏｏｅｇｎｇｍｅｈｄｗｉｈｎｎ—ｉｕｏｄａｉｎｌｓｐｒｐｓｄ．Ｉｈｓｍｅｈｄ，ｃｏｓｃｒｅａｉｎｂｔｅｈｅｅｖｄｓｇｎｔｉｔｏｒｓ — ｏｒｌｔｏｅｗｅｎｔｅｒｃｉｅｉ—

ACOUSTIC UNDERWATER ANTENNA

专利名称：ACOUSTIC UNDERWATER ANTENNA 发明人：PETER, HOFFMANN申请号：EP07818953.7申请日：20071012公开号：EP2087378A2公开日：20090812专利内容由知识产权出版社提供摘要：In an acoustic underwater antenna which has electroacoustic transducers (29) which are arranged one after another at fixed distances from one another, electronic assemblies associated with the transducers (29), for signal processing of the electrical transducer signals, and an antenna casing (28) surrounding the transducers (29) and the electronic assemblies, in order to achieve a transducer separation which can be reduced increasingly in order to achieve a high transmission and/or reception frequency for the underwater antenna for example, two electronic assemblies of adjacent transducers (29) in the antenna casing (28) are in each case combined to form an electronic module (28) arranged between the transducers (29), and one of the two adjacent transducers is in each case arranged on mutually averted, free module faces.申请人：ATLAS ELEKTRONIK GMBH地址：Sebaldsbrücker Heerstrasse 235 28309 Bremen DE国籍：DE代理机构：Wasiljeff, Johannes M.B.更多信息请下载全文后查看。

jstd035声学扫描

JOINT INDUSTRY STANDARDAcoustic Microscopy for Non-HermeticEncapsulatedElectronicComponents IPC/JEDEC J-STD-035APRIL1999Supersedes IPC-SM-786 Supersedes IPC-TM-650,2.6.22Notice EIA/JEDEC and IPC Standards and Publications are designed to serve thepublic interest through eliminating misunderstandings between manufacturersand purchasers,facilitating interchangeability and improvement of products,and assisting the purchaser in selecting and obtaining with minimum delaythe proper product for his particular need.Existence of such Standards andPublications shall not in any respect preclude any member or nonmember ofEIA/JEDEC or IPC from manufacturing or selling products not conformingto such Standards and Publications,nor shall the existence of such Standardsand Publications preclude their voluntary use by those other than EIA/JEDECand IPC members,whether the standard is to be used either domestically orinternationally.Recommended Standards and Publications are adopted by EIA/JEDEC andIPC without regard to whether their adoption may involve patents on articles,materials,or processes.By such action,EIA/JEDEC and IPC do not assumeany liability to any patent owner,nor do they assume any obligation whateverto parties adopting the Recommended Standard or ers are alsowholly responsible for protecting themselves against all claims of liabilities forpatent infringement.The material in this joint standard was developed by the EIA/JEDEC JC-14.1Committee on Reliability Test Methods for Packaged Devices and the IPCPlastic Chip Carrier Cracking Task Group(B-10a)The J-STD-035supersedes IPC-TM-650,Test Method2.6.22.For Technical Information Contact:Electronic Industries Alliance/ JEDEC(Joint Electron Device Engineering Council)2500Wilson Boulevard Arlington,V A22201Phone(703)907-7560Fax(703)907-7501IPC2215Sanders Road Northbrook,IL60062-6135 Phone(847)509-9700Fax(847)509-9798Please use the Standard Improvement Form shown at the end of thisdocument.©Copyright1999.The Electronic Industries Alliance,Arlington,Virginia,and IPC,Northbrook,Illinois.All rights reserved under both international and Pan-American copyright conventions.Any copying,scanning or other reproduction of these materials without the prior written consent of the copyright holder is strictly prohibited and constitutes infringement under the Copyright Law of the United States.IPC/JEDEC J-STD-035Acoustic Microscopyfor Non-Hermetic EncapsulatedElectronicComponentsA joint standard developed by the EIA/JEDEC JC-14.1Committee on Reliability Test Methods for Packaged Devices and the B-10a Plastic Chip Carrier Cracking Task Group of IPCUsers of this standard are encouraged to participate in the development of future revisions.Contact:EIA/JEDEC Engineering Department 2500Wilson Boulevard Arlington,V A22201 Phone(703)907-7500 Fax(703)907-7501IPC2215Sanders Road Northbrook,IL60062-6135 Phone(847)509-9700Fax(847)509-9798ASSOCIATION CONNECTINGELECTRONICS INDUSTRIESAcknowledgmentMembers of the Joint IPC-EIA/JEDEC Moisture Classiﬁcation Task Group have worked to develop this document.We would like to thank them for their dedication to this effort.Any Standard involving a complex technology draws material from a vast number of sources.While the principal members of the Joint Moisture Classiﬁcation Working Group are shown below,it is not possible to include all of those who assisted in the evolution of this Standard.To each of them,the mem-bers of the EIA/JEDEC and IPC extend their gratitude.IPC Packaged Electronic Components Committee ChairmanMartin FreedmanAMP,Inc.IPC Plastic Chip Carrier Cracking Task Group,B-10a ChairmanSteven MartellSonoscan,Inc.EIA/JEDEC JC14.1CommitteeChairmanJack McCullenIntel Corp.EIA/JEDEC JC14ChairmanNick LycoudesMotorolaJoint Working Group MembersCharlie Baker,TIChristopher Brigham,Hi/FnRalph Carbone,Hewlett Packard Co. Don Denton,TIMatt Dotty,AmkorMichele J.DiFranza,The Mitre Corp. Leo Feinstein,Allegro Microsystems Inc.Barry Fernelius,Hewlett Packard Co. Chris Fortunko,National Institute of StandardsRobert J.Gregory,CAE Electronics, Inc.Curtis Grosskopf,IBM Corp.Bill Guthrie,IBM Corp.Phil Johnson,Philips Semiconductors Nick Lycoudes,MotorolaSteven R.Martell,Sonoscan Inc. Jack McCullen,Intel Corp.Tom Moore,TIDavid Nicol,Lucent Technologies Inc.Pramod Patel,Advanced Micro Devices Inc.Ramon R.Reglos,XilinxCorazon Reglos,AdaptecGerald Servais,Delphi Delco Electronics SystemsRichard Shook,Lucent Technologies Inc.E.Lon Smith,Lucent Technologies Inc.Randy Walberg,NationalSemiconductor Corp.Charlie Wu,AdaptecEdward Masami Aoki,HewlettPackard LaboratoriesFonda B.Wu,Raytheon Systems Co.Richard W.Boerdner,EJE ResearchVictor J.Brzozowski,NorthropGrumman ES&SDMacushla Chen,Wus Printed CircuitCo.Ltd.Jeffrey C.Colish,Northrop GrummanCorp.Samuel J.Croce,Litton AeroProducts DivisionDerek D-Andrade,Surface MountTechnology CentreRao B.Dayaneni,Hewlett PackardLaboratoriesRodney Dehne,OEM WorldwideJames F.Maguire,Boeing Defense&Space GroupKim Finch,Boeing Defense&SpaceGroupAlelie Funcell,Xilinx Inc.Constantino J.Gonzalez,ACMEMunir Haq,Advanced Micro DevicesInc.Larry A.Hargreaves,DC.ScientiﬁcInc.John T.Hoback,Amoco ChemicalCo.Terence Kern,Axiom Electronics Inc.Connie M.Korth,K-Byte/HibbingManufacturingGabriele Marcantonio,NORTELCharles Martin,Hewlett PackardLaboratoriesRichard W.Max,Alcatel NetworkSystems Inc.Patrick McCluskey,University ofMarylandJames H.Moffitt,Moffitt ConsultingServicesRobert Mulligan,Motorola Inc.James E.Mumby,CibaJohn Northrup,Lockheed MartinCorp.Dominique K.Numakura,LitchﬁeldPrecision ComponentsNitin B.Parekh,Unisys Corp.Bella Poborets,Lucent TechnologiesInc.D.Elaine Pope,Intel Corp.Ray Prasad,Ray Prasad ConsultancyGroupAlbert Puah,Adaptec Inc.William Sepp,Technic Inc.Ralph W.Taylor,Lockheed MartinCorp.Ed R.Tidwell,DSC CommunicationsCorp.Nick Virmani,Naval Research LabKen Warren,Corlund ElectronicsCorp.Yulia B.Zaks,Lucent TechnologiesInc.IPC/JEDEC J-STD-035April1999 iiTable of Contents1SCOPE (1)2DEFINITIONS (1)2.1A-mode (1)2.2B-mode (1)2.3Back-Side Substrate View Area (1)2.4C-mode (1)2.5Through Transmission Mode (2)2.6Die Attach View Area (2)2.7Die Surface View Area (2)2.8Focal Length(FL) (2)2.9Focus Plane (2)2.10Leadframe(L/F)View Area (2)2.11Reﬂective Acoustic Microscope (2)2.12Through Transmission Acoustic Microscope (2)2.13Time-of-Flight(TOF) (3)2.14Top-Side Die Attach Substrate View Area (3)3APPARATUS (3)3.1Reﬂective Acoustic Microscope System (3)3.2Through Transmission AcousticMicroscope System (4)4PROCEDURE (4)4.1Equipment Setup (4)4.2Perform Acoustic Scans..........................................4Appendix A Acoustic Microscopy Defect CheckSheet (6)Appendix B Potential Image Pitfalls (9)Appendix C Some Limitations of AcousticMicroscopy (10)Appendix D Reference Procedure for PresentingApplicable Scanned Data (11)FiguresFigure1Example of A-mode Display (1)Figure2Example of B-mode Display (1)Figure3Example of C-mode Display (2)Figure4Example of Through Transmission Display (2)Figure5Diagram of a Reﬂective Acoustic MicroscopeSystem (3)Figure6Diagram of a Through Transmission AcousticMicroscope System (3)April1999IPC/JEDEC J-STD-035iiiIPC/JEDEC J-STD-035April1999This Page Intentionally Left BlankivApril1999IPC/JEDEC J-STD-035 Acoustic Microscopy for Non-Hermetic EncapsulatedElectronic Components1SCOPEThis test method deﬁnes the procedures for performing acoustic microscopy on non-hermetic encapsulated electronic com-ponents.This method provides users with an acoustic microscopy processﬂow for detecting defects non-destructively in plastic packages while achieving reproducibility.2DEFINITIONS2.1A-mode Acoustic data collected at the smallest X-Y-Z region deﬁned by the limitations of the given acoustic micro-scope.An A-mode display contains amplitude and phase/polarity information as a function of time ofﬂight at a single point in the X-Y plane.See Figure1-Example of A-mode Display.IPC-035-1 Figure1Example of A-mode Display2.2B-mode Acoustic data collected along an X-Z or Y-Z plane versus depth using a reﬂective acoustic microscope.A B-mode scan contains amplitude and phase/polarity information as a function of time ofﬂight at each point along the scan line.A B-mode scan furnishes a two-dimensional(cross-sectional)description along a scan line(X or Y).See Figure2-Example of B-mode Display.IPC-035-2 Figure2Example of B-mode Display(bottom half of picture on left)2.3Back-Side Substrate View Area(Refer to Appendix A,Type IV)The interface between the encapsulant and the back of the substrate within the outer edges of the substrate surface.2.4C-mode Acoustic data collected in an X-Y plane at depth(Z)using a reﬂective acoustic microscope.A C-mode scan contains amplitude and phase/polarity information at each point in the scan plane.A C-mode scan furnishes a two-dimensional(area)image of echoes arising from reﬂections at a particular depth(Z).See Figure3-Example of C-mode Display.1IPC/JEDEC J-STD-035April1999IPC-035-3 Figure3Example of C-mode Display2.5Through Transmission Mode Acoustic data collected in an X-Y plane throughout the depth(Z)using a through trans-mission acoustic microscope.A Through Transmission mode scan contains only amplitude information at each point in the scan plane.A Through Transmission scan furnishes a two-dimensional(area)image of transmitted ultrasound through the complete thickness/depth(Z)of the sample/component.See Figure4-Example of Through Transmission Display.IPC-035-4 Figure4Example of Through Transmission Display2.6Die Attach View Area(Refer to Appendix A,Type II)The interface between the die and the die attach adhesive and/or the die attach adhesive and the die attach substrate.2.7Die Surface View Area(Refer to Appendix A,Type I)The interface between the encapsulant and the active side of the die.2.8Focal Length(FL)The distance in water at which a transducer’s spot size is at a minimum.2.9Focus Plane The X-Y plane at a depth(Z),which the amplitude of the acoustic signal is maximized.2.10Leadframe(L/F)View Area(Refer to Appendix A,Type V)The imaged area which extends from the outer L/F edges of the package to the L/F‘‘tips’’(wedge bond/stitch bond region of the innermost portion of the L/F.)2.11Reflective Acoustic Microscope An acoustic microscope that uses one transducer as both the pulser and receiver. (This is also known as a pulse/echo system.)See Figure5-Diagram of a Reﬂective Acoustic Microscope System.2.12Through Transmission Acoustic Microscope An acoustic microscope that transmits ultrasound completely through the sample from a sending transducer to a receiver on the opposite side.See Figure6-Diagram of a Through Transmis-sion Acoustic Microscope System.2April1999IPC/JEDEC J-STD-0353IPC/JEDEC J-STD-035April1999 3.1.6A broad band acoustic transducer with a center frequency in the range of10to200MHz for subsurface imaging.3.2Through Transmission Acoustic Microscope System(see Figure6)comprised of:3.2.1Items3.1.1to3.1.6above3.2.2Ultrasonic pulser(can be a pulser/receiver as in3.1.1)3.2.3Separate receiving transducer or ultrasonic detection system3.3Reference packages or standards,including packages with delamination and packages without delamination,for use during equipment setup.3.4Sample holder for pre-positioning samples.The holder should keep the samples from moving during the scan and maintain planarity.4PROCEDUREThis procedure is generic to all acoustic microscopes.For operational details related to this procedure that apply to a spe-ciﬁc model of acoustic microscope,consult the manufacturer’s operational manual.4.1Equipment Setup4.1.1Select the transducer with the highest useable ultrasonic frequency,subject to the limitations imposed by the media thickness and acoustic characteristics,package conﬁguration,and transducer availability,to analyze the interfaces of inter-est.The transducer selected should have a low enough frequency to provide a clear signal from the interface of interest.The transducer should have a high enough frequency to delineate the interface of interest.Note:Through transmission mode may require a lower frequency and/or longer focal length than reﬂective mode.Through transmission is effective for the initial inspection of components to determine if defects are present.4.1.2Verify setup with the reference packages or standards(see3.3above)and settings that are appropriate for the trans-ducer chosen in4.1.1to ensure that the critical parameters at the interface of interest correlate to the reference standard uti-lized.4.1.3Place units in the sample holder in the coupling medium such that the upper surface of each unit is parallel with the scanning plane of the acoustic transducer.Sweep air bubbles away from the unit surface and from the bottom of the trans-ducer head.4.1.4At aﬁxed distance(Z),align the transducer and/or stage for the maximum reﬂected amplitude from the top surface of the sample.The transducer must be perpendicular to the sample surface.4.1.5Focus by maximizing the amplitude,in the A-mode display,of the reﬂection from the interface designated for imag-ing.This is done by adjusting the Z-axis distance between the transducer and the sample.4.2Perform Acoustic Scans4.2.1Inspect the acoustic image(s)for any anomalies,verify that the anomaly is a package defect or an artifact of the imaging process,and record the results.(See Appendix A for an example of a check sheet that may be used.)To determine if an anomaly is a package defect or an artifact of the imaging process it is recommended to analyze the A-mode display at the location of the anomaly.4.2.2Consider potential pitfalls in image interpretation listed in,but not limited to,Appendix B and some of the limita-tions of acoustic microscopy listed in,but not limited to,Appendix C.If necessary,make adjustments to the equipment setup to optimize the results and rescan.4April1999IPC/JEDEC J-STD-035 4.2.3Evaluate the acoustic images using the failure criteria speciﬁed in other appropriate documents,such as J-STD-020.4.2.4Record the images and theﬁnal instrument setup parameters for documentation purposes.An example checklist is shown in Appendix D.5IPC/JEDEC J-STD-035April19996April1999IPC/JEDEC J-STD-035Appendix AAcoustic Microscopy Defect Check Sheet(continued)CIRCUIT SIDE SCANImage File Name/PathDelamination(Type I)Die Circuit Surface/Encapsulant Number Affected:Average%Location:Corner Edge Center (Type II)Die/Die Attach Number Affected:Average%Location:Corner Edge Center (Type III)Encapsulant/Substrate Number Affected:Average%Location:Corner Edge Center (Type V)Interconnect tip Number Affected:Average%Interconnect Number Affected:Max.%Length(Type VI)Intra-Laminate Number Affected:Average%Location:Corner Edge Center Comments:CracksAre cracks present:Yes NoIf yes:Do any cracks intersect:bond wire ball bond wedge bond tab bump tab leadDoes crack extend from leadﬁnger to any other internal feature:Yes NoDoes crack extend more than two-thirds the distance from any internal feature to the external surfaceof the package:Yes NoAdditional veriﬁcation required:Yes NoComments:Mold Compound VoidsAre voids present:Yes NoIf yes:Approx.size Location(if multiple voids,use comment section)Do any voids intersect:bond wire ball bond wedge bond tab bump tab lead Additional veriﬁcation required:Yes NoComments:7IPC/JEDEC J-STD-035April1999Appendix AAcoustic Microscopy Defect Check Sheet(continued)NON-CIRCUIT SIDE SCANImage File Name/PathDelamination(Type IV)Encapsulant/Substrate Number Affected:Average%Location:Corner Edge Center (Type II)Substrate/Die Attach Number Affected:Average%Location:Corner Edge Center (Type V)Interconnect Number Affected:Max.%LengthLocation:Corner Edge Center (Type VI)Intra-Laminate Number Affected:Average%Location:Corner Edge Center (Type VII)Heat Spreader Number Affected:Average%Location:Corner Edge Center Additional veriﬁcation required:Yes NoComments:CracksAre cracks present:Yes NoIf yes:Does crack extend more than two-thirds the distance from any internal feature to the external surfaceof the package:Yes NoAdditional veriﬁcation required:Yes NoComments:Mold Compound VoidsAre voids present:Yes NoIf yes:Approx.size Location(if multiple voids,use comment section)Additional veriﬁcation required:Yes NoComments:8Appendix BPotential Image PitfallsOBSERV ATIONS CAUSES/COMMENTSUnexplained loss of front surface signal Gain setting too lowSymbolization on package surfaceEjector pin knockoutsPin1and other mold marksDust,air bubbles,ﬁngerprints,residueScratches,scribe marks,pencil marksCambered package edgeUnexplained loss of subsurface signal Gain setting too lowTransducer frequency too highAcoustically absorbent(rubbery)ﬁllerLarge mold compound voidsPorosity/high concentration of small voidsAngled cracks in package‘‘Dark line boundary’’(phase cancellation)Burned molding compound(ESD/EOS damage)False or spotty indication of delamination Low acoustic impedance coating(polyimide,gel)Focus errorIncorrect delamination gate setupMultilayer interference effectsFalse indication of adhesion Gain set too high(saturation)Incorrect delamination gate setupFocus errorOverlap of front surface and subsurface echoes(transducerfrequency too low)Fluidﬁlling delamination areasApparent voiding around die edge Reﬂection from wire loopsIncorrect setting of void gateGraded intensity Die tilt or lead frame deformation Sample tiltApril1999IPC/JEDEC J-STD-0359Appendix CSome Limitations of Acoustic MicroscopyAcoustic microscopy is an analytical technique that provides a non-destructive method for examining plastic encapsulated components for the existence of delaminations,cracks,and voids.This technique has limitations that include the following: LIMITATION REASONAcoustic microscopy has difficulty inﬁnding small defects if the package is too thick.The ultrasonic signal becomes more attenuated as a function of two factors:the depth into the package and the transducer fre-quency.The greater the depth,the greater the attenuation.Simi-larly,the higher the transducer frequency,the greater the attenu-ation as a function of depth.There are limitations on the Z-axis(axial)resolu-tion.This is a function of the transducer frequency.The higher the transducer frequency,the better the resolution.However,the higher frequency signal becomes attenuated more quickly as a function of depth.There are limitations on the X-Y(lateral)resolu-tion.The X-Y(lateral)resolution is a function of a number of differ-ent variables including:•Transducer characteristics,including frequency,element diam-eter,and focal length•Absorption and scattering of acoustic waves as a function of the sample material•Electromechanical properties of the X-Y stageIrregularly shaped packages are difficult to analyze.The technique requires some kind ofﬂat reference surface.Typically,the upper surface of the package or the die surfacecan be used as references.In some packages,cambered packageedges can cause difficulty in analyzing defects near the edgesand below their surfaces.Edge Effect The edges cause difficulty in analyzing defects near the edge ofany internal features.IPC/JEDEC J-STD-035April1999 10April1999IPC/JEDEC J-STD-035Appendix DReference Procedure for Presenting Applicable Scanned DataMost of the settings described may be captured as a default for the particular supplier/product with speciﬁc changes recorded on a sample or lot basis.Setup Conﬁguration(Digital Setup File Name and Contents)Calibration Procedure and Calibration/Reference Standards usedTransducerManufacturerModelCenter frequencySerial numberElement diameterFocal length in waterScan SetupScan area(X-Y dimensions)Scan step sizeHorizontalVerticalDisplayed resolutionHorizontalVerticalScan speedPulser/Receiver SettingsGainBandwidthPulseEnergyRepetition rateReceiver attenuationDampingFilterEcho amplitudePulse Analyzer SettingsFront surface gate delay relative to trigger pulseSubsurface gate(if used)High passﬁlterDetection threshold for positive oscillation,negative oscillationA/D settingsSampling rateOffset settingPer Sample SettingsSample orientation(top or bottom(ﬂipped)view and location of pin1or some other distinguishing characteristic) Focus(point,depth,interface)Reference planeNon-default parametersSample identiﬁcation information to uniquely distinguish it from others in the same group11IPC/JEDEC J-STD-035April1999Appendix DReference Procedure for Presenting Applicable Scanned Data(continued) Reference Procedure for Presenting Scanned DataImageﬁle types and namesGray scale and color image legend deﬁnitionsSigniﬁcance of colorsIndications or deﬁnition of delaminationImage dimensionsDepth scale of TOFDeviation from true aspect ratioImage type:A-mode,B-mode,C-mode,TOF,Through TransmissionA-mode waveforms should be provided for points of interest,such as delaminated areas.In addition,an A-mode image should be provided for a bonded area as a control.12Standard Improvement FormIPC/JEDEC J-STD-035The purpose of this form is to provide the Technical Committee of IPC with input from the industry regarding usage of the subject standard.Individuals or companies are invited to submit comments to IPC.All comments will be collected and dispersed to the appropriate committee(s).If you can provide input,please complete this form and return to:IPC2215Sanders RoadNorthbrook,IL 60062-6135Fax 847509.97981.I recommend changes to the following:Requirement,paragraph number Test Method number,paragraph numberThe referenced paragraph number has proven to be:Unclear Too RigidInErrorOther2.Recommendations forcorrection:3.Other suggestions for document improvement:Submitted by:Name Telephone Company E-mailAddress City/State/ZipDate ASSOCIATION CONNECTING ELECTRONICS INDUSTRIESASSOCIATION CONNECTINGELECTRONICS INDUSTRIESISBN#1-580982-28-X2215 Sanders Road, Northbrook, IL 60062-6135Tel. 847.509.9700 Fax 847.509.9798。

UNDERWATER EXTERIOR SHIP HULL IMAGING SYSTEM EMPLO

专利名称：UNDERWATER EXTERIOR SHIP HULLIMAGING SYSTEM EMPLOYING A REMOTEMICROPROCESSOR CONTROLLED ACOUSTICTRANSDUCER ARRAY发明人：Mark S. Rogers申请号：US10862028申请日：20040604公开号：US20060114748A1公开日：20060601专利内容由知识产权出版社提供专利附图：摘要：A multi-beam acoustic transducer array system for producing color enhanced,three dimensional, high resolution images of a ship's underwater hull, wherein the acoustic transducers can be mounted in orthogonal pairs, each pair being positioned opposite from another pair within a shipping channel. The array can be utilized either in a stationary configuration within a controlled shipping lane or suspended in the water column from a mobile support vessel, with at least one array being orthogonally mounted and suspended in the water column from the mobile support vessel, wherein the mobile configuration obviates the need for multiple orthogonal arrays, for purposes of performing acquisition and imaging of a ship's hull. Each orthogonal array consists of a first transducer transmitting sonar pulses along a horizontal plane and a second transducer transmitting sonar pulses along a vertical plane, such that the two beaming sonar pulses are orthogonal, thereby providing optimal coverage.申请人：Mark S. Rogers地址：St. Petersburg FL US国籍：US更多信息请下载全文后查看。

Experimental Demonstration of Underwater Acoustic

Experimental Demonstration of Underwater Acoustic Communication by Vector SensorsAijun Song,Member,IEEE,Ali Abdi,Senior Member,IEEE,Mohsen Badiey,Member,IEEE,and Paul HurskyAbstract—Acoustic communication often relies on a large size array with multiple spatially separated hydrophones to deal with the challenging underwater channel.This poses limitation to its application in compact size underwater platforms.In this paper, acoustic communication by vector sensors is demonstrated by the data collected during a high frequency acoustic experiment,where a vector sensor array was drifting in the ocean.It is shown that the multichannel receiver using a single vector sensor can offer sig-niﬁcant size reduction for coherent acoustic communication at the carrier frequency of12kHz,compared with a pressure sensor line array.Further,the performance difference between vector sen-sors and pressure sensors varies at communication ranges.At close ranges(up to160m),both a single vector sensor and a vector sensor array can offer signiﬁcant performance gain compared with the pressure sensor array.At longer ranges(up to1080m),the single vector sensor has the same performance with the pressure sensor array,on average.The vector sensor array consistently provides gain at all ranges over the pressure sensor array since additional information of the acousticﬁeld is utilized by vector sensors. Index Terms—Underwater acoustic communication,time re-versal,vector sensor.I.I NTRODUCTIONU NDERWATER acoustic communication is critical to a number of civilian and scientiﬁc missions in the ocean, for example navigation and communication for underwater au-tonomous vehicles(AUVs).However,the underwater channel is challenging for digital communication[1].To deal with signiﬁcant multipath and fast channelﬂuctuations,reception diversity from spatially separated hydrophones is often em-ployed to achieve acceptable performance.Readers can refer to [2]–[4]for multichannel decision feedback equalization(DFE) and[5]–[10]for time reversal methods as design examples. These designs often lead to the usage of a large size array, which might be impossible to accommodate at compact under-Manuscript received June09,2010;revised January15,2011;accepted March16,2011.Date of publication June16,2011;date of current version July 01,2011.This work was supported by the Ofﬁce of Naval Research(ONR) Code321OA(Grant N00014-01-1-0114)and the National Science Foundation (NSF)(Grant CCF-0830204and CCF-0830190).Part of the content has been presented at the OCEANS’08meeting,Quebec,QC,Canada,September 15–18,2008.Associate Editor:S.Zhou.A.Song and M.Badiey are with College of Earth,Ocean,and Environment, University of Delaware,Newark,DE19716USA(e-mail:ajsong@). Ali Abdi is with Center for Wireless Communications and Signal Processing Research,Electrical and Computer Engineering Department,New Jersey Insti-tute of Technology,Newark,NJ07102USA.Pual Hursky is with Heat,Light,and Sound Research,Inc.,La Jolla,CA 92037USA.Color versions of one or more of theﬁgures in this paper are available online at .Digital Object Identiﬁer10.1109/JOE.2011.2133050water platforms such as AUVs.Moreover,smaller array size is always preferred in underwater missions because it leads to easier operations.In this paper,we explore the possibilities of using vector sensors as an alternative to spatially separated pressure sensors.Acoustic vector sensors are capable of measuring three orthogonal particle velocity components of the acousticﬁeld, in addition to the scalar acoustic pressure,at a collocated point in space[11].They have long been considered for localization and detection of underwater objects[11]–[15].For example, four physical quantities measured by a vector sensor have been processed by a multichannelﬁlter to perform beamforming [16]–[18].Vector sensors and vector sensor arrays can offer improved performance in direction-ﬁnding applications com-pared with their scalar pressure counterparts[18].It has been proposed in[19]that a single vector can serve as a multichannel receiver for underwater digital communication through the use of both scalar pressure and particle velocity components of the acousticﬁeld.Multiuser communication with vector sensors has been investigated through computer simulations in[20]. Motivated by the potential applications of vector sensors in underwater communication,statistical correlation models for acoustic vector sensor arrays have been developed in[21].In this paper,we use the experimental data to demonstrate the usefulness of velocity channels for acoustic communication. Both a single vector sensor and an array of vector sensors are investigated for coherent acoustic communication using exper-imental data.The paper is organized as follows.System equa-tions for communication by vector sensors are brieﬂy reviewed in Section II.Channel properties and computer simulations for the usage of vector sensors in coherent underwater communica-tion are discussed in Section III.Experimental demonstration of the concept is provided in Section IV and concluding remarks are given in Section V.II.S YSTEM E QUATIONSIn this section,basic system equations for data detection via a vector sensor introduced in[22]are brieﬂy reviewed.To demon-strate the basic concepts of how both the vector and scalar com-ponents of the acousticﬁeld can be utilized for data reception, we consider a simple system with one transducer and one vector sensor.As shown in Fig.1,the vector sensor denoted as a black square measures the pressure and the,and components of the particle velocity.A.Pressure and Velocity Channels and NoiseThere are four channels in Fig.1:the pressure channel,rep-resented by a straight dashed line,and three pressure-equiva-0364-9059/$26.00©2011IEEElent velocity channels ,and ,shown by curved dashed lines.To de ﬁne ,and ,we ﬁrst de ﬁne the particle ve-locities,,and .According to the linearized equation for time-harmonic waves,the ,and components of the ve-locity at the frequency are given by [23](1)where is the density of the ﬂuid and .Equation (1)states that the velocity in a certain direction is proportional to the spatial pressure gradient in that direction [23].The as-sociated pressure-equivalent velocity channels are de ﬁned as,and ,which give(2)where is the acoustic wavenumber and is the sound speed.The additive ambient noise pressure at the receiver is shown by in Fig.1.Similar to (1),the ,and components of ambient noise velocities are,and ,respec-tively.So we can obtain the pressure-equivalent ambient noisevelocities as(3)B.Input–Output System EquationsAccording to Fig.1,the received pressure signal in response to the signal transmitted from the transmitter can be written as .Here stands for convolution in time and is the pressure channel impulse response.We also de-ﬁne the pressure-equivalent received velocity signals as,and .Based on (2)and by taking the spatial gradient with respect to,and axes,we obtain key system equations(4)Note that the four output signals are measured at a collocated point in space.With the assumption that the noise is spher-ically isotropic,the noise terms in (4)are uncorrelated [14].In addition to the noise correlation property,the arrival struc-ture and correlation functions of the channels are relevant to acoustic communication performance.As shown by the numer-ical acoustic simulations in [19],[22],the pressure channel and the velocity channels can provide a new form of diversity,sim-ilar to an array of spatially separated pressure sensors.There-fore,the pressure source and vector sensor in Fig.1have the po-Fig.1.A 14vector sensor communication system with a sound source anda vector sensor receiver in the underwater environment.The vector sensor mea-sures pressure and,and components of the acoustic particle velocity,all at a single point in space.tential to form a single-input–multiple-output (SIMO)system.In the next section,relevant channel characteristics are reported including delay spread,noise correlation,and channel correla-tion from experimental data.III.M EASURED C HANNEL C HARACTERISTICSDuring a high frequency Makai acoustic communication ex-periment (MakaiEx)conducted west off the Kauai Island,HI,in September and October of 2005[24],a ﬁve element Wilcoxon array was deployed multiple times.The measured characteris-tics of particle velocity channels are presented in this section.The performance of particle velocity channels in digital commu-nication is investigated by computer simulations to gain some initial insight in this section.Performance obtained from at-sea experiments will be discussed in Section IV.A.Channel Measurements During MakaiExEach element of the Wilcoxon vector sensor array had three velocity-meters that were sensitive only along a speci ﬁc direc-tion,besides an embedded omnidirectional pressure sensor [15].Therefore,each vector sensor had four channels and generated four data streams:one pressure channel and three ,and components of the particle velocity.The length of each vector sensor was 6.6cm,and the element spacing (center-to-center distance)was 10cm.In one deployment on September 23,2005,a bottom mounted acoustic source continuously transmitted a series of communication signals at the carrier frequency of12kHz.The water depth at the source was about 100m and the source depth was about 95m.The vector sensor array was at-tached to A-frame steel cable of the drifting R/V Kilo Moana .The array was considered vertical since a 200pound weight was attached to the end of the array cable.The top element is referred to as the ﬁrst.The fourth element,referred to as the bottom ele-ment,was about 40m below the sea surface.The ﬁfth element did not function properly during the experiment.During the ex-periment,the R/V Kilo Moana was drifting in deeper water.As mentioned,a pressure source and a vector sensor can form a 14SIMO communication system,similar to four pressure channels of four vector sensors.Note the four channels of a vector sensor are colocated at a single point in space,whereas the four pressure channels have an aperture of 30cm.In what follows,channel characteristics and receiver performance of these two systems are compared.Fig.2shows an example of the impulse response functions obtained from the ﬁeld data.AFig.2.Normalized amplitudes of the measured impulse responses(a)from the pressure channel,-velocity,-velocity,and-velocity channels of theﬁrst sensor; and(b)from the four pressure channels of the array.pseudorandom binary phase-shift keying(BPSK)signal at the carrier frequency of12kHz was used to probe the acoustic channel during the experiment.The symbol rate of the BPSK signal was6kilosymbols/s and the utilized bandwidth was6 kHz.The impulse response was obtained by the least squares channel estimation algorithm.For this particular data set,the source–receiver range was about20m.The range was calcu-lated based on the ship GPS data and source position.As shown in Fig.2(a),the-,-,and-particle velocity channels had an arrival structure similar to the pressure channel.However,later arrivals of and particle velocity channels were weaker.This resulted in smaller root-mean-square(RMS)delay spreads,de-ﬁned as[25](5) where is a discrete,baseband impulse response sampled at a period of.Speciﬁcally,the delay spread of the pressure channel was3.9ms,whereas those of the,,and chan-nels were4.6,2.0,and2.6ms,respectively.Considering the 30-cm aperture of the pressure sensor line array,the four pres-sure channel impulse responses showed similarity among them-selves,as demonstrated in Fig.2(b).The delay spread of the four pressure channels were3.7,4.3,5.2,and4.6ms,respectively. Because a small delay spread corresponds to less intersymbol interference(ISI),the-and-velocity channels might offer better communication results than the pressure channels[19]. Besides the delay spread,channel/noise correlations are also relevant to the receiver performance.Table I shows the noise and channel correlation among the multiple channels of the vector sensor and of the pressure sensor line array.The correlation numbers in Table I are the modulus of the complex correlation deﬁned as(6)TABLE IC HANNEL AND N OISE C ORRELATION M EASUREDF ROM THE F IELD DATA where are two complex sequences and represents the expectation operation.The channel impulse responses shown in Fig.2were used for the correlation calculation.For the vector sensor,the pressure,-velocity,-velocity,and-velocity chan-nels of theﬁrst sensor were numbered as channel#1to channel #4,respectively.The pressure outputs of the four vector sen-sors were used in the correlation analysis of the pressure sensor array.The noise correlation calculation used3.75s ambient noise,which was recorded20s before the BPSK training se-quence.As shown in Table I,although the four channels of the vector sensor were colocated at a single point,correlation among some of the channels could be small.Further,most of the noise corre-lation numbers of the vector sensor were smaller than those of the pressure sensor array.B.Simulated Performance of Vector Sensor ReceiversIn Fig.3,the bit error rates(BERs)of the receivers using a single vector sensor,a single pressure sensor,and a four-element pressure sensor array are shown.The experimental impulse re-sponses shown in Fig.2were used.The classic symbol-spaced DFE was used to compensate for the ISI in a single channel asFig.3.Simulated performance of the receivers using a single vector sensor,a four-element pressure sensor array,and a single pressure sensor.The impulse responses of Fig.2were used.The size of the four element pressure sensor array was30cm,whereas the vector sensor size was6.6cm.well as in multiple channels[2],[26].Carrier phase offset and time variation of the channel were not considered.The number of the feedforward taps was for the multichannel DFE and the number of the feedback taps was149sym-bols.The BERs corresponded to a6kb/s uncoded BPSK data stream at12kHz.As expected,the-velocity channel receiver had about1-dB reduction in the required signal-to-noise ratio(SNR)over the single pressure sensor receiver for the BER of.This was attributed to the smaller delay spread of the-velocity channel.By using all the channels of the vector sensor,a7-dB reduction in the required SNR for the BER of could be obtained,compared with a single pressure sensor receiver.The vector sensor receiver showed modest improve-ment(1.5dB)over the four-channel pressure sensor array.The average SNR of each multichannel receiver is deﬁned as[19](7) where,and are the average powers of the pres-sure channel and of the particle velocity channels,and ,and are their respective noise powers.The vector sensor had a compact size(6.6cm)compared with the30-cm pressure sensor array.This is crucial in acoustic modem applications for AUVs where there are serious limita-tions on receiver size.This beneﬁt is the result of the colocated particle velocity information that can be measured by a compact vector sensor.This is an alternative to spatially separated pres-sure sensors to achieve reception diversity.IV.C OMMUNICATION E XPERIMENT U SING V ECTOR S ENSORS In this section,we present a practical receiver to utilize par-ticle velocity channels in the underwater environment.Experi-mental data collected at multiple communication ranges are an-alyzed.Performance comparison between vector sensors and a pressure sensor array is reported as well.A.Multichannel Equalizer to Utilizethe Particle Velocity ChannelsConsidering the time-varying,dispersive properties of the un-derwater channel,(4)can be rewritten for a receiver with multiple vector sensors asto(8) where the pressure channel,-velocity channel,-velocity channel,and-velocity channel of the th vector sensor, ,correspond to channel indexesto.In(8),is the instantaneous carrier phase offset.,is the discrete-time baseband impulse response function where is the impulse response length in symbols.Note that(8)can also be used as the input–output equation for an-element pressure sensor array.Fig.4shows the receiver structure,which is similar to the algorithm presented in[4].As a channel-estimation-based processor,the multichannel equalizer tracks channel and carrier phase variations to accommodate fast channel variations.The channel estimate can be obtained from the received signal and the past decision results using the least squares method[27].The receiver consists of four compo-nents:noise normalization,phase tracking and correction,time reversal multichannel combining,andﬁnally a single channel DFE.A noise normalization component is needed since the noise power usually is not uniformly distributed among the pressure channel and the velocity channels[14].The noise power in the channel estimator[27]can be used as an estimate of the noise power(9)Fig.4.The practical multichannel equalizer consists of four parts:noise normalization,phase tracking and correction,time reversal multichannel combining,and single-channel DFE.where is the source symbol of the preamble and is channel estimation block size.After noise normalization,time reversal multichannel com-bining uses time reversed channel estimates as matched-ﬁlters for the phase-corrected signals on each channel[28].The output after time reversal combining is(10) where is the combined noise term and is the ef-fective impulse response between the source and output of time reversal.After time reversal,a single channel DFE with joint phase tracking[2]is used to compensate for the residual ISI and phase ﬂuctuations in.The exponentially weighted recursive least-squares(RLS)algorithm is used to update the equalizer tap weights.The mean squared error(MSE)at the soft output of the DFE,,and the BER of the hard decision,,are used as performance metrics in the next subsection.The receiver in Fig.4is equivalent to the optimal multi-channel DFE structure[2].Rather than applying feedforward ﬁlters to the individual channels,the proposed receiver uses a single channel DFE on the composite channel after time reversal combining.The proposed receiver avoids,explicit or inexplicit, calculation and updating of large number of feedforwardﬁlters for individual channels.The length of the single channel DFE ﬁlters in the proposed receiver are shorter since the effective impulse response is usually compact and near time-in-variant.This lead to low complexity of the proposed receiver in the time-varying,dispersive underwater channel.munication Data AnalysisIn Section III,the impulse responses at the communication range of20m were presented for the particle velocity channels. In this section,communication data from seven source–receiver ranges up to1km are analyzed.A uniform set of receiver param-eters were used to demodulate all the BPSK data including the measurements from pressure channels as well as from particle velocity channels.The received data were oversampled with an oversampling rate of.The estimated channel length was25ms or150symbols.The channel estimation block size was chosen to be three times of the channel length,i.e., .The number of the feedforward taps wasfor the fractionally spaced DFE,where was the feed-forwardﬁlter span in symbols.The number of the feedback taps was5symbols.At the beginning of the3.75-s BPSK packet,1600symbols were used for noise normalization, initial channel estimation,phase tracking,and DFE tap weight training.The RLS forgetting factor in the DFE was0.999. For seven BPSK transmissions at different ranges,the demod-ulation results are shown in Table II.The time reversal receiver in Fig.4was applied to four channels of a single vector sensor, the four pressure channels of the four vector sensors(a pressure sensor array),and16channels of the four vector sensors(a vector sensor array).The receiver used previous detected symbols for channel estimation and equalization after the preamble.It is noted that there was no difference in the observed phase offset between the particle velocity channel and the pressure channels.This is consistent with the theoretical analysis in[29],where it is shown that Doppler spreads of particle velocity and pressure channels are nearly the same.When there were excessive errors for a de-modulation block,for example BER greater than0.3,the receiver would not be able to estimate the channel.We then consider this as receiver failure as marked in Table II for several cases.The av-erage channel delay spread and average correlation coefﬁcients are listed for each range in Table III.The average was calculated over the3.75-s BPSK packet.The results can be discussed into three range groups.Theﬁrst group includes the close range(20,80,and160m).The re-ceivers using vector sensors had signiﬁcant performance advan-tage at these close communication ranges.For example,at the 20-and80-m range,the single vector sensor receiver had about 5.5-and2.5-dB reduction in the output MSE over the pressure sensor receiver.At160m,the pressure sensor array receiver failed whereas the single vector sensor had an output MSE of 4.7dB.TABLE IID EMODULATIONR ESULTS D URING M AKAI EXTABLE IIIM EASURED C HANNEL C HARACTERISTICS D URING M AKAI EXAt these ranges,the direct path was strong and multipath ef-fect was limited due to the close range.The channels were very similar to impulse responses shown in Fig.2.For the vector sensor,the -and -velocity channels had weaker later arrivals re ﬂected as smaller delay spread of -and -velocity channels in Table III for these ranges.The performance gain was partially attributed to the smaller delay spread of the -and -velocity channels.For the second range group (290and 430m),the pressure sensor array outperformed the single vector sensor.The receiver with the single vector sensor failed at the 290-m range while the output MSE of the pressure sensor array was 4.8dB.The per-formance gain of the pressure sensor array was partly due to lower correlation among its four pressure elements than among the four channels of the vector sensor.At the 430-m range,the pressure sensor array had modest gain over the single vector sensor,1.4-dB reduction in the output MSE.Note that such gain of the pressure sensor array at these ranges came with the price of the larger size receiving array,compared with a compact vector sensor [22].At these ranges,the channels had strong mul-tipath and later arrivals ﬂuctuated very fast.As an example,the pressure and velocity channels of a vector sensor at the 430-m range are shown in Fig.5.The velocity channels did not have the advantage of smaller delay spread or lower correlation.How-ever,the vector sensor array had about 1.4-dB improvements over the pressure sensor array in terms of the output MSE.For the third range group (900and 1080m),the single vector sensor performed better than the pressure sensor array:the output MSE was 4.6and 4.1dB,respectively for these two ranges,for the single vector sensor while the receiver with the pressure sensor array failed.This was due to the high correlation (close to 0.9and higher)among all the channels of the pressure sensor array at these two ranges,as listed in Table III.The vector sensor array had similar performance to the single vector sensor,con ﬁrming strong correlation among the corresponding channels from multiple elements of the array.V .C ONCLUSIONIn this paper,particle velocity channels provided by vector sensors were utilized for underwater acoustic communication.Through the use of the MakaiEx data,coherent communication by vector sensors was demonstrated between a bottom mounted sound source and a drifting vector sensor vertical array.A prac-tical multichannel equalizer with near-optimal performance was implemented to process the data at multiple communica-tion ranges.Channel parameters such as correlation and delay spread that affect data communication were calculated from the measurements.It was shown that the multichannel equalizer using a single vector sensor can offer signi ﬁcant receiver size reduction for coherent acoustic communication at the carrier frequency of 12kHz,compared with a 30-cm aperture pressure sensor line array.Further,the performance difference betweenFig.5.The measured3.75-s impulse responses of(a)pressure channel,(b)-velocity,(c)-velocity,and(d)-velocity channels at the source–receiver range of about430m.vector sensors and pressure sensors varied at communication ranges.At close ranges(up to160m),both a single vector sensor and a vector sensor array offered signiﬁcant performance gain compared with the pressure sensor array.At longer ranges (up to1080m),the vector sensor array provided consistent performance gain,although smaller than those at close ranges, over the pressure sensor array since additional information of the acousticﬁeld was utilized.These results suggest that vector sensors can offer acoustic communication solutions that are particularly needed in the compact underwater platforms.A CKNOWLEDGMENTThe authors would like to thank all the participants of MakaiEx.They also would like to give special thanks to B. Abraham(Applied Physical Sciences),who participated in the vector sensor experiment as part of MakaiEx.R EFERENCES[1]D.B.Kilfoyle and A.B.Baggeroer,“The state of the art in underwateracoustic telemetry,”IEEE J.Ocean.Eng.,vol.25,no.1,pp.4–27,Jan.2000.[2]M.Stojanovic,J.Catipovic,and J.G.Proakis,“Adaptive multichannelcombining and equalization for underwater acoustic communications,”J.Acoust.Soc.Amer.,vol.94,no.3,pp.1621–1631,Sep.1993. [3]M.Stojanovic,J.Catipovic,and J.G.Proakis,“Reduced-complexityspatial and temporal processing of underwater acoustic communication signals,”J.Acoust.Soc.Amer.,vol.94,no.2,pp.961–972, Aug.1995.[4]A.Song,M.Badiey,H.-C.Song,W.S.Hodgkiss,and M.B.Porter,TheKauaiEx Group,“Impact of ocean variability on coherent underwater acoustic communications during the Kauai experiment(KauaiEx),”J.Acoust.Soc.Amer.,vol.123,no.2,pp.856–865,Feb.2008.[5]G.F.Edelmann,T.Akal,W.S.Hodgkiss,S.Kim,W.A.Kuperman,and H.C.Song,“An initial demonstration of underwater acoustic com-munications using time reversal,”IEEE J.Ocean.Eng.,vol.31,no.3, pp.602–609,Jul.2002.[6]D.Rouseff,D.R.Jackson,W.L.J.Fox,C.D.Jones,J.A.Ritcey,and D.R.Dowling,“Underwater acoustic communication by passive-phase conjugation:Theory and experimental results,”IEEE J.Ocean.Eng.,vol.26,no.4,pp.821–831,Oct.2001.[7]T.C.Yang,“Temporal resolutions of time-reversal and passive-phaseconjugation for underwater acoustic communications,”IEEE J.Ocean.Eng.,vol.28,no.2,pp.229–245,Apr.2003.[8]G.F.Edelmann,H.C.Song,S.Kim,W.S.Hodgkiss,W.A.Ku-perman,and T.Akal,“Underwater acoustic communications using time reversal,”IEEE J.Ocean.Eng.,vol.30,no.4,pp.852–864,Oct.2005.[9]T.C.Yang,“Correlation-based decision-feedback equalizer for under-water acoustic communications,”IEEE J.Ocean.Eng.,vol.30,no.4, pp.865–880,Oct.2005.[10]H.C.Song,W.S.Hodgkiss,W.A.Kuperman,M.Stevenson,and T.Akal,“Improvement of time reversal communications using adaptive channel equalizers,”IEEE J.Ocean.Eng.,vol.31,no.2,pp.487–496, Apr.2006.[11]A.Nehorai and E.Paldi,“Acoustic vector-sensor array processing,”IEEE Trans.Signal Process.,vol.42,no.9,pp.2481–2491,Sep.1994.[12]K.T.Wong and M.D.Zoltowski,“Closed-form underwater acousticdirection-ﬁnding with arbitrarily spaced vector-hydrophones at un-known locations,”IEEE J.Ocean.Eng.,vol.22,no.3,pp.566–575, Jul.1997.[13]M.Hawkes and A.Nehorai,“Acoustic vector-sensor beamforming andCapon direction estimation,”IEEE Trans.Signal Process.,vol.46,no.9,pp.2291–2304,Sep.1998.[14]M.Hawkes and A.Nehorai,“Acoustic vector-sensor correlations inambient noise,”IEEE J.Ocean.Eng.,vol.26,no.3,pp.337–347,Jul.2001.[15]J.C.Shipps and B.M.Abraham,“The use of vector sensors for un-derwater port and waterway security,”in Proc.ISA/IEEE Sensors Ind.Conf.,New Orleans,LA,2004.[16]G.L.D’Spain,W.S.Hodgkiss,and G.L.Edmonds,“The simultaneousmeasurement of infrasonic acoustic particle velocity and acoustic pres-sure in the ocean by freely drifting swallowﬂoats,”IEEE J.Ocean.Eng.,vol.16,no.2,pp.195–207,Apr.2001.[17]B.A.Cray and A.H.Nuttall,“Directivity factors for linear arrays ofvelocity sensors,”J.Acoust.Soc.Amer.,vol.110,no.1,pp.324–331, Jul.2001.[18]G.L.D’Spain,J.C.Luby,G.R.Wilson,and R.A.Gramann,“Vectorsensors and vector sensor line arrays:Comments on optimal array gain and detection,”J.Acoust.Soc.Amer.,vol.120,no.1,pp.171–185,Jul.2006.[19]A.Abdi,H.Guo,and P.Sutthiwan,“A new vector sensor receiverfor underwater acoustic communication,”in Proc.Oceans Conf.,Van-couver,BC,Canada,2007.[20]H.Guo and A.Abdi,“Multiuser underwater communication withspace-time block codes and acoustic vector sensors,”in Proc.Oceans Conf.,Quebec City,QC,Canada,2008.[21]A.Abdi and H.Guo,“Signal correlation modeling in acoustic vectorsensor arrays,”IEEE Trans.Signal Process.,vol.57,no.3,pp.892–903,Mar.2009.[22]A.Abdi and H.Guo,“A new compact multichannel receiver forunderwater wireless communication networks,”IEEE Trans.Wireless Commun.,vol.8,no.7,pp.3326–3329,Jul.2009.[23]A.D.Pierce,Acoustics:An Introduction to Its Physical Principles andApplications,2nd ed.New York:AIP,1989,pp.14–20.[24]M.B.Porter et al.,“The Makai experiment:High frequency acoustics,”in Proc.Eur.Conf.Underwater Acoust.,Carvoeiro,Portugal,2006.[25]J.D.Parsons,The Mobile Radio Propagation Channel,4th ed.NewYork:Wiley,1992,pp.164–189.[26]J.G.Proakis,Digital Communications,4th ed.New York:McGraw-Hill,2000,pp.638–642.[27]J.A.Flynn,J.A.Ritcey,D.Rouseff,and W.L.J.Fox,“Multichannelequalization by decision-directed passive phase conjugation:Experi-mental results,”IEEE J.Ocean.Eng.,vol.29,no.3,pp.824–836,Jul.2004.[28]W.A.Kuperman,W.S.Hodgkiss,H.C.Song,T.Akal,C.Ferla,andD.R.Jackson,“Phase conjugation in the ocean:Experimental demon-stration of an acoustic time-reversal mirror,”J.Acoust.Soc.Amer.,vol.103,no.1,pp.25–40,Jan.1998.[29]H.Guo,A.Abdi,A.Song,and M.Badiey,“Delay and Dopplerspreads in underwater acoustic particle velocity channels,”J.Acoust.Soc.Amer.,vol.129,no.4,pp.2015–2025,Apr.2011.Aijun Song(S’02–M’05)received the Ph.D.de-gree in electrical engineering at the University of Delaware,Newark,in2005.From2005to2008,he was a Postdoctoral Re-search Associate at the College of Earth,Ocean, and Environment(CEOE),University of Delaware. During this period,he was also an Ofﬁce of Naval Research(ONR)Postdoctoral Fellow supported by the special research award from the Ocean Acoustics program.Since2008,he has been an Assistant Re-search Professor of the Physical Ocean Science and Engineering program,University of Delaware.His general interests include underwater acoustic signal propagation,digital communication theory,and advanced signal processing in mobile radio frequency and underwater acousticenvironments.Ali Abdi(S’98–M’01–SM’06)received the Ph.D.degree in electrical engineering from the Universityof Minnesota,Minneapolis.He joined the Department of Electrical and Com-puter Engineering of New Jersey Institute of Tech-nology(NJIT),Newark,where he is currently an As-sociate Professor.His current research interests in-clude digital communication and propagation mod-eling in underwater and terrestrial channels,charac-terization and estimation of wireless channels,blindmodulation recognition and interference cancellation techniques.Dr.Abdi was an Associate Editor for IEEE T RANSACTIONS ON V EHICULAR T ECHNOLOGY from2002to2007.His most recent professional activity was co-organizing and co-chairing a full-day special session entitled“Acoustic Par-ticle Velocity and Vector Fields:Signal Processing and Communication Appli-cations,”for the159th Meeting of the Acoustical Society of America in2010. His recent awards include2008New Jersey Inventors Hall of Fame(NJIHoF) Innovators Award and2009IEEE Region1Award for outstanding leadership and contributions in underwatercommunication.Mohsen Badiey(M’94)received the Ph.D.degree inapplied marine physics and ocean engineering fromthe University of Miami,Rosenstiel School of Ma-rine and Atmospheric Science,Miami,FL,in1988.He was a Postdoctoral Fellow at the Port and Har-bour Institute,Ministry of Transport,Tokyo,Japan,from1988through1990,and worked on the waterwave interaction with seaﬂoor and seismic wavepropagation.In1990,he became a faculty memberat the College of Earth,Ocean,and Environment,University of Delaware,where he is presently a full Professor and Director of Physical Ocean Science and Engineering program. From1992to1995,he directed the ocean acoustics program at the Ofﬁce of Naval Research.Since the beginning of2000,he has been one of the Principal Investigators to establish a long-term program in studying high frequency acoustic waves in relationship to underwater acoustic communications,a research program that resulted in Kauai Experiment in2003and its follow upﬁeld measurements(Makai Experiment,2005and KAM08Experiment, 2008).His research interests are physics of sound and vibration,underwater acoustics in shallow water regions,acoustical oceanography,underwater acoustic communications,seabed acoustics,and geophysics.Dr.Badiey is a Fellow of the Acoustical Society ofAmerica.Paul Hursky received the B.S.and M.S.degrees inphysics from the University of Pennsylvania in1978and1980,and the Ph.D.degree in electrical engi-neering from University of California,San Diego in2001.From1980to1999,he was with Lockheed Aero-nautical Systems Company,in Burbank and SanDiego,CA,where he worked on signal processingalgorithm development for sonar systems.From1999to2004,he was with Science ApplicationsInternational Inc.,where he worked on advanced concepts for underwater acoustic applications.From2004to the present,he has been with Heat,Light,and Sound Research Inc.,a company he helped start.He now works on research topics in underwater acoustics,including acoustic prop-agation modeling,model-based signal processing,marine mammal tracking, and acoustic communications.Dr.Hursky is a member of the IEEE Signal Processing,Ocean Engineering, Communications,and Information Theory Societies,and the Acoustical Society of America.。

水下目标的Gammatone子带降噪和希尔伯特-黄变换特征提取

水下目标的Gammatone子带降噪和希尔伯特-黄变换特征提取王曙光;曾向阳;王征;王强【摘要】水下目标识别是水声探测中的关键技术,具有重要的应用价值.海洋环境的复杂性导致水下目标识别中存在不可回避的噪声干扰.以人耳听觉机理为基础,提出了一种结合Gammatone滤波器、小波阈值降噪和希尔伯特-黄变换(HHT)的水下目标识别方法.采用Gammatone滤波器实现人耳听觉机理的模拟,并在此基础上进行小波阈值降噪,提高系统的噪声鲁棒性,然后利用HHT进行时频分析和特征提取.利用实际水下目标数据进行识别实验,对提出的方法进行了验证.实验结果表明,提出的方法在低信噪比条件下具有良好的鲁棒性,并具有较好的识别效果.【期刊名称】《兵工学报》【年(卷),期】2015(036)009【总页数】6页(P1704-1709)【关键词】声学;水下目标识别;人耳听觉;Gammatone滤波器;希尔伯特-黄变换【作者】王曙光;曾向阳;王征;王强【作者单位】西北工业大学航海学院,陕西西安710072;西北工业大学航海学院,陕西西安710072;中国船舶工业集团公司,北京100048;西北工业大学航海学院,陕西西安710072【正文语种】中文【中图分类】TN912.34水下目标识别是水声探测领域的关键技术之一，是探测系统智能化的重要标志。

在现代海战中，能否对水下目标实现快速、准确的探测和分类，对海军舰艇的作战能力和生存能力至关重要。

海洋环境复杂、多变的特性对水下目标识别系统的精确度和稳定性都提出了巨大的挑战。

海洋环境噪声和海洋混响是水下目标识别系统的主要干扰，如果能够有效地消除这些噪声的影响，便有望建立具有高精度的鲁棒性水下目标识别系统。

人耳听觉系统在嘈杂的环境下仍可以有效地跟踪并识别感兴趣的声音，而不受其他噪声的干扰，即所谓的“鸡尾酒会效应”。

这种优秀的识别能力是许多机器识别方法不能比拟的。

人耳听觉的一个显著特点是非线性，其对声音的感知可以划分为一系列不同带宽的频带，通过对各频带的感知实现对声信号的整体把握。

光照和噪声条件下的人脸识别

光照和噪声条件下的人脸识别
杜平;徐大为;刘重庆
【期刊名称】《上海交通大学学报》
【年(卷),期】2003(37)9
【摘要】讨论了光照和噪声条件下的人脸识别问题,提出了基于Gabor滤波器对人脸图像提取Gabor特征并与EFM+MAP-2相结合的人脸识别方法.实验证明,该方法在非均匀光照和噪声条件下比其他传统的人脸识别方法具有更好的鲁棒性.【总页数】5页(P1443-1446)
【关键词】人脸识别;Gabor滤波器;增强型线性判别模型;最大后验概率
【作者】杜平;徐大为;刘重庆
【作者单位】上海交通大学图像处理与模式识别研究所
【正文语种】中文
【中图分类】TP391.41
【相关文献】
1.不同光照条件下CCD相机时间噪声和空间噪声的研究 [J], 韩采芹;李华;朱顺一;沈忙作
2.探究遮挡和复杂光照条件下鲁棒人脸识别算法 [J], 尚晓锐
3.探究遮挡和复杂光照条件下鲁棒人脸识别算法 [J], 尚晓锐
4.光照变化条件下的人脸识别技术探讨 [J], 周令
5.一种复杂光照条件下人脸识别签到系统的设计 [J], 周敏维
因版权原因，仅展示原文概要，查看原文内容请购买。