基于FPGA语音识别系统设计与实现
合集下载
- 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)。
广西师范大学 硕士学位论文 基于FPGA语音识别系统设计与实现 姓名:王明娟 申请学位级别:硕士 专业:电路与系统 指导教师:胡维平
20090401
广西师范大学硕士学位论文
基于 FPGA 语音识别系统设计与实现
研究生:王明娟
导师:胡维平 教授
Baidu Nhomakorabea
专业:电路与系统
研究方向:语音信号处理
年级:2006级
摘要
The system adopts sopc technology based on NIOS II,whole design is based on DE2 development platform.The advantages of using this solution is the implementation of on chip system and reduction of physical volume as well as the total power consumption;meanwhile the system control core realizes within the FPGA device, it could be renewal and updated conveniently,so it increases the generality and maintainability vary much.moreover,because the system needs plenty of highspeed calculation,through the design,the author makes full use of abundant hardware multiplier in Cyclone II chips,and realizes endpoint detection module of speech signal, FFT rapid fourier transform module,DCT discrete cosine transform module and so on.In order to improve whole performance of the system,the author makes full use of highspeed parallel processing advantage of FPGA and
本论文主要包含了以下几个方面: (1)结合ALTERA CYCLONE II芯片的特点,确定了基于FPGA语音识别系统的总体设计, 在此基础上进行了系统的软硬件的选择和设计。 (2)自主设计了纯硬件描述语言的驱动电路设计,完成了高速语音采集的工作,并 且对存储数据芯片SRAM中的原始语音数据进行提取导入MATLAB平台测试数据的正确性。整 个程序测试的方式对系统的模块测试起到重要的作用。 (3)完成高速定点256点的FFT模块的设计,此模块是系统成败的关键,实现高速实 时的运算。 (4)结合SOPC的特性,设计了人机友好接口,如LCD显示屏的提示反馈信息等等,以 及利用ALTERA提供的一些驱动接口设计完成用户定制的系统。 (5)进行了整体系统测试,系统可以较稳定地实现实时处理的目的,具有一定的市 场潜在价值。 【关键词】FPGA,SOPC,语音识别,语音采集,FFT IP核
论文使用授权声明
本人完全了解广西师范大学有关保留、使用学位论文的规定。广 西师范大学、中国科学技术信息研究所、清华大学论文合作部,有权 保留本人所送交学位论文的复印件和电子文档,可以采用影印、缩印 或其他复制手段保存论文。本人电子文档的内容和纸质论文的内容相 一致。除在保密期内的保密论文外,允许论文被查阅和借阅,可以公 布(包括刊登)论文的全部或部分内容。论文的公布(包括刊登)授 权广西师范大学学位办办理。
III
论文独创性声明
本人郑重声明:所提交的学位论文是本人在导师的指导下进行的 研究工作及取得的成果。除文中已经注明引用的内容外,本论文不含 其他个人或其他机构已经发表或撰写过的研究成果。对本文的研究作 出重要贡献的个人和集体,均已在文中以明确方式标明。本人承担本 声明的法律责任。
研究生签名:
日期:
研究生签名: 导 师签名:
日期: 日期:
II
广西师范大学硕士学位论文
custom-defined hardware components which hangs on the Avalon BUS in supporting development environment,enable the system to process digital signal efficiently, meanwhile its performance is better than traditional microcontroller and common DSP chips.
This paper mainly includes the following aspects: (1)In combination with the characteristics of ALTERA CYCLONE II chips,author determined the overall design of speech recognition system Based on FPGA,and carried out systematic software and hardware selection and design on this basis. (2)Author independent designed driving circuit design in the pure hardware description language,accomplishing the highspeed speech acquisition,meanwhile extracting the speech data from SRAM and introcucing into MATLAB platform for test and measurement.The way of program testing played an important role to systematic module test. ( 3 ) Author completd the design of highspeed fixed-point 256 points FFT module,which is the key of success or failure throughout the system,realizing highspeed real-time operation. (4)Author designed man-machine interface with SOPC characteristics,such as suggesting feedback imformation on LCD,designed and finished custom-defined system which can be offered by ALTERA’s drivie interfaces. (5)Author made the whole systematic measurement,the system can quite stablely realize real-time processing,and has certain market potential value. 【Key word】FPGA,SOPC,speech recognition,speech acquisition,FFT IP core
近年来,语音识别研究大部分集中在算法设计和改进等方面,而随着半导体技术的高 速发展,集成电路规模的不断增大与各种研发技术水平的不断提高,新的硬件平台的推出, 语音识别实现平台有了更多的选择。语音识别技术在与DSP、FPGA、ASIC等器件为平台的 嵌入式系统结合后,逐渐向实用化、小型化方向发展。
本课题通过对现有各种语音特征参数与孤立词语音识别模型进行研究的基础上,重点 探索基于动态时间规整算法的DTW模型在孤立词语音识别领域的应用,并结合基于FPGA的 SOPC系统,在嵌入式平台上实现具有较好精度与速度的孤立词语音识别系统。
I
广西师范大学硕士学位论文
Design and Implementation of Speech Recognition
System Based on FPGA
Candidate :Wang Ming-jian Advisor : Hu Wei-ping professor Speciality :Circuit and System Research Direction :speech signal processing Grade :2006
Abstract
In the recent years,research for speech recognition mostly focuses on algorithm design and improvement.With the rapid development of semiconductor technology, scale with integrated circuit continuously enlarging, and level of all kinds of development technique continuously improving,new hardware platform protruding, there are a lot of selections on speech recognition platform. Combination with DSP、FPGA、ASIC devices which are embedded systems,speech recognition research gradually looks toward to practical development and miniaturization. Based on the research of all kinds of existing speech feature parameters and isolated words speech recognition models,the topic mainly explores the application of DTW model based on dynamic programming in the field of isolated words speech recognition, and combining with SOPC system based on FPGA, realizes a good speech recognition system with accuracy and speed.
本系统整体设计基于DE2开发平台,采用基于Nios II的SOPC技术。采用这种解决方案 的优点是实现了片上系统,减少了系统的物理体积和总体功耗;同时系统控制核心都在 FPGA内部实现,可以极为方便地更新和升级系统,大大地提高了系统的通用性和可维护性。 此外,由于本系统需要大量的高速数据运算,在设计中作者充分利用了Cyclone II 芯片 的丰富的硬件乘法器,实现了语音信号的端点检测模块,FFT快速傅立叶变换模块,DCT离 散余弦变换模块等硬件设计模块。为了提高系统的整体性能,作者充分利用了FPGA的高速 并行的优势,以及配套开发环境中的Avalon总线自定义硬件外设,使系统处理数字信号的 能力大大提高,其性能优于传统的微控制器和普通DSP芯片。
20090401
广西师范大学硕士学位论文
基于 FPGA 语音识别系统设计与实现
研究生:王明娟
导师:胡维平 教授
Baidu Nhomakorabea
专业:电路与系统
研究方向:语音信号处理
年级:2006级
摘要
The system adopts sopc technology based on NIOS II,whole design is based on DE2 development platform.The advantages of using this solution is the implementation of on chip system and reduction of physical volume as well as the total power consumption;meanwhile the system control core realizes within the FPGA device, it could be renewal and updated conveniently,so it increases the generality and maintainability vary much.moreover,because the system needs plenty of highspeed calculation,through the design,the author makes full use of abundant hardware multiplier in Cyclone II chips,and realizes endpoint detection module of speech signal, FFT rapid fourier transform module,DCT discrete cosine transform module and so on.In order to improve whole performance of the system,the author makes full use of highspeed parallel processing advantage of FPGA and
本论文主要包含了以下几个方面: (1)结合ALTERA CYCLONE II芯片的特点,确定了基于FPGA语音识别系统的总体设计, 在此基础上进行了系统的软硬件的选择和设计。 (2)自主设计了纯硬件描述语言的驱动电路设计,完成了高速语音采集的工作,并 且对存储数据芯片SRAM中的原始语音数据进行提取导入MATLAB平台测试数据的正确性。整 个程序测试的方式对系统的模块测试起到重要的作用。 (3)完成高速定点256点的FFT模块的设计,此模块是系统成败的关键,实现高速实 时的运算。 (4)结合SOPC的特性,设计了人机友好接口,如LCD显示屏的提示反馈信息等等,以 及利用ALTERA提供的一些驱动接口设计完成用户定制的系统。 (5)进行了整体系统测试,系统可以较稳定地实现实时处理的目的,具有一定的市 场潜在价值。 【关键词】FPGA,SOPC,语音识别,语音采集,FFT IP核
论文使用授权声明
本人完全了解广西师范大学有关保留、使用学位论文的规定。广 西师范大学、中国科学技术信息研究所、清华大学论文合作部,有权 保留本人所送交学位论文的复印件和电子文档,可以采用影印、缩印 或其他复制手段保存论文。本人电子文档的内容和纸质论文的内容相 一致。除在保密期内的保密论文外,允许论文被查阅和借阅,可以公 布(包括刊登)论文的全部或部分内容。论文的公布(包括刊登)授 权广西师范大学学位办办理。
III
论文独创性声明
本人郑重声明:所提交的学位论文是本人在导师的指导下进行的 研究工作及取得的成果。除文中已经注明引用的内容外,本论文不含 其他个人或其他机构已经发表或撰写过的研究成果。对本文的研究作 出重要贡献的个人和集体,均已在文中以明确方式标明。本人承担本 声明的法律责任。
研究生签名:
日期:
研究生签名: 导 师签名:
日期: 日期:
II
广西师范大学硕士学位论文
custom-defined hardware components which hangs on the Avalon BUS in supporting development environment,enable the system to process digital signal efficiently, meanwhile its performance is better than traditional microcontroller and common DSP chips.
This paper mainly includes the following aspects: (1)In combination with the characteristics of ALTERA CYCLONE II chips,author determined the overall design of speech recognition system Based on FPGA,and carried out systematic software and hardware selection and design on this basis. (2)Author independent designed driving circuit design in the pure hardware description language,accomplishing the highspeed speech acquisition,meanwhile extracting the speech data from SRAM and introcucing into MATLAB platform for test and measurement.The way of program testing played an important role to systematic module test. ( 3 ) Author completd the design of highspeed fixed-point 256 points FFT module,which is the key of success or failure throughout the system,realizing highspeed real-time operation. (4)Author designed man-machine interface with SOPC characteristics,such as suggesting feedback imformation on LCD,designed and finished custom-defined system which can be offered by ALTERA’s drivie interfaces. (5)Author made the whole systematic measurement,the system can quite stablely realize real-time processing,and has certain market potential value. 【Key word】FPGA,SOPC,speech recognition,speech acquisition,FFT IP core
近年来,语音识别研究大部分集中在算法设计和改进等方面,而随着半导体技术的高 速发展,集成电路规模的不断增大与各种研发技术水平的不断提高,新的硬件平台的推出, 语音识别实现平台有了更多的选择。语音识别技术在与DSP、FPGA、ASIC等器件为平台的 嵌入式系统结合后,逐渐向实用化、小型化方向发展。
本课题通过对现有各种语音特征参数与孤立词语音识别模型进行研究的基础上,重点 探索基于动态时间规整算法的DTW模型在孤立词语音识别领域的应用,并结合基于FPGA的 SOPC系统,在嵌入式平台上实现具有较好精度与速度的孤立词语音识别系统。
I
广西师范大学硕士学位论文
Design and Implementation of Speech Recognition
System Based on FPGA
Candidate :Wang Ming-jian Advisor : Hu Wei-ping professor Speciality :Circuit and System Research Direction :speech signal processing Grade :2006
Abstract
In the recent years,research for speech recognition mostly focuses on algorithm design and improvement.With the rapid development of semiconductor technology, scale with integrated circuit continuously enlarging, and level of all kinds of development technique continuously improving,new hardware platform protruding, there are a lot of selections on speech recognition platform. Combination with DSP、FPGA、ASIC devices which are embedded systems,speech recognition research gradually looks toward to practical development and miniaturization. Based on the research of all kinds of existing speech feature parameters and isolated words speech recognition models,the topic mainly explores the application of DTW model based on dynamic programming in the field of isolated words speech recognition, and combining with SOPC system based on FPGA, realizes a good speech recognition system with accuracy and speed.
本系统整体设计基于DE2开发平台,采用基于Nios II的SOPC技术。采用这种解决方案 的优点是实现了片上系统,减少了系统的物理体积和总体功耗;同时系统控制核心都在 FPGA内部实现,可以极为方便地更新和升级系统,大大地提高了系统的通用性和可维护性。 此外,由于本系统需要大量的高速数据运算,在设计中作者充分利用了Cyclone II 芯片 的丰富的硬件乘法器,实现了语音信号的端点检测模块,FFT快速傅立叶变换模块,DCT离 散余弦变换模块等硬件设计模块。为了提高系统的整体性能,作者充分利用了FPGA的高速 并行的优势,以及配套开发环境中的Avalon总线自定义硬件外设,使系统处理数字信号的 能力大大提高,其性能优于传统的微控制器和普通DSP芯片。