語音識(shí)別技術(shù)及應(yīng)用

全文43頁 約18000字 論述翔實(shí)

摘要
語音信號(hào)處理的主要技術(shù)有語音壓縮、識(shí)別、合成等，我所研究的課題就是其中一項(xiàng)——語音識(shí)別。語音識(shí)別技術(shù)是信息領(lǐng)域的標(biāo)志性技術(shù)，隨著計(jì)算機(jī)技術(shù)的飛速發(fā)展，其技術(shù)日臻成熟，目前正處于向產(chǎn)品化邁進(jìn)的轉(zhuǎn)折階段，它作為人機(jī)對(duì)話的手段，在計(jì)算機(jī)日益普及的今天，愈發(fā)顯現(xiàn)出其在IT產(chǎn)業(yè)中的重要地位。
論文的內(nèi)容主要有三部分：第一部分闡述了語音識(shí)別技術(shù)的發(fā)展歷史與現(xiàn)狀以及原理和方法，并對(duì)幾種方法進(jìn)行了比較。第二部分根據(jù)對(duì)理論的研究選擇了合適的端點(diǎn)檢測(cè)方法、特征參數(shù)提取以及識(shí)別算法，其中，端點(diǎn)檢測(cè)算法選擇雙門限檢測(cè)方法，特征參數(shù)選擇MFCC；由于是特定人識(shí)別，匹配算法選擇模板匹配的代表——DTW算法。并在此基礎(chǔ)上設(shè)計(jì)了一款聲控電視遙控器，畫出了硬件及軟件結(jié)構(gòu)框圖。第三部分通過編程對(duì)所選擇的主要識(shí)別算法進(jìn)行了MATLAB仿真，所實(shí)現(xiàn)的識(shí)別率基本達(dá)到了要求。

［關(guān)鍵詞］語音識(shí)別，DTW，MFCC，遙控器，MATALB

ABSTRACT

Voice compression, speech recognition and speech synthesis are The primary kinds of technology of speech signal processing. And what I research for is just a kind of them— speech recognition. It is the symbol technology in the field of communication. The technology becomes more and more mature as the computer technology develops fast, and it’s on the way to become product. Nowadays, computer is more and more common, so as a method of man-machine conversation, it certainly shows more important position in IT industry.
The paper consists of three parts. In the first one, I have expound the history, actuality, principles and methods of speech recognition, and also compared some of the methods. In the second part, I choose appropriate measures of endpoint detection, feature selection and also recognition. Thereinto, the choice of the first one is double-limit detection, the feature is MFCC; because it is talker-dependent recognition, I choose the representation of templates matching—DTW. Basing on all of these, I design a kind of remote device which is controlled by speech, in which both the hardware and software configuration are shown. In the last segment, I workout emulators with MATLAB, which are according to the recognition methods chosen above. And the probability of correct recognition is satisfying by and large.

[key Words] speech recognition，DTW，MFCC，remote device，MATALB

目錄

1緒論 3
1.1 語音識(shí)別技術(shù)的發(fā)展歷史 3
1.2 語音識(shí)別技術(shù)的應(yīng)用 4
1.3 設(shè)計(jì)背景及主要研究?jī)?nèi)容 7
2語音識(shí)別基礎(chǔ) 8
2.1 語音信號(hào)處理的技術(shù)基礎(chǔ) 8
2.2 語音識(shí)別的基本原理 9
2.2.1 預(yù)處理 10
2.2.2 特征提取 10
2.2.3 語音模型庫 11
2.2.4 模式匹配 11
2.2.5 后處理 11
2.3 語音識(shí)別的分類 12
2.4 語音識(shí)別存在的問題 13
2.5 語音識(shí)別的優(yōu)點(diǎn) 13
3語音識(shí)別的基本方法 14
3.1 語音識(shí)別的一般方法 14
3.2 模板匹配法 18
3.3 DTW算法 19
4聲控電視遙控器的設(shè)計(jì) 20
4.1 遙控器的發(fā)展歷史及工作原理 20
4.2 聲控遙控器的總體設(shè)計(jì) 21
4.2.1系統(tǒng)硬件總體設(shè)計(jì)思路 21
4.2.2按鍵及功能簡(jiǎn)圖 24
4.2.3紅外發(fā)射電路 24
4.2.4系統(tǒng)軟件系統(tǒng)的總體設(shè)計(jì) 25
5系統(tǒng)仿真 27
5.1 端點(diǎn)檢測(cè)方法 27
5.2 特征參數(shù)提取 31
5.3 匹配算法——DTW算法 34
5.4 總體仿真結(jié)果 37
6總結(jié)與展望 38
參考文獻(xiàn) 40
附錄1 42

部分參考文獻(xiàn)

【21】 Lawrence Rabiner,Biing-Hwang．Fundamentals of Speech Recognition．北京:清華大學(xué)出版社,1999
【22】 楊行峻．語音信號(hào)數(shù)字處理．北京：電子工業(yè)出版社，1995
【23】 廣州迅控電子科技有限公司材料
【24】 李向陽、朱學(xué)峰．家用通用遙控器的開發(fā)．華南理工大學(xué)學(xué)報(bào)，1999．8
【25】 李晶皎．嵌入式語音技術(shù)及凌陽16位單片機(jī)應(yīng)用．北京航空航天大學(xué)出版社,2003