臺灣博碩士論文加值系統

本論文中，利用隱藏式馬可夫模型建立語音模型，針對國語字音來做辨識，研究可分為：語音訊號特徵值之擷取、碼書的取得、隱藏式馬可夫模型及語言文法知識的使用；以往大多數的研究都只偏向語音訊號特徵方面，而忽略了語言文法知識對語音辨識的重要性。

This thesis describes a Chinese speech recognition system.
The system uses cepstrum-coefficient, vector quantization, Hidden M-arkov model and grammar knowledge.The majority of previous resear-ch projects tend to study the characteristics of voice signals, and ignore the importance of grammar knowledge for voice recognition.

中文摘要......................................................Ⅰ
英文摘要......................................................Ⅱ
誌謝..........................................................Ⅲ
目錄..........................................................Ⅳ
表目錄........................................................Ⅵ
圖目錄........................................................Ⅶ一、緒論.......................................................1
1.1研究動機與目的..............................1
1.2研究方法....................................1
1.3各章節內容概述..............................2
二、語音訊號辨識的理論基礎....................................4
2.1前言.................................................4
2.2語音訊號前置處理.....................................5
2.2.1 能量量測................................................5
2.2.2 越零率..................................................6
2.2.3 語音信號切割............................................6
2.3 語音訊號的特徵及特徵參數及抽取............................8
2.3.1 將語音分成一序列的音框.........................8
2.3.2 高頻濾波器的處理...............................9
2.3.3 將各個音框乘上漢明視窗..............................9
2.3.4 替每一個音框求出一組線性預測係數....................9
2.3.5 求倒頻譜參數.......................................11
2.4 向量量化.................................................12
2.5 隱藏式馬可夫模型之建立...................................13
2.5.1 機率計算......................................13
2.5.2 正算程序............................................14
2.5.3 逆算程序............................................16
2.6 參數重估..................................................18
2.7 實作HMMs..................................................20
2.7.1 Scaling ......................................20
2.7.2 多個觀測序列..................................21
三、語音訊號辨識系統的建立...................................22
3.1語音訊號特徵參數抽取與特徵參數的量化.......22
3.2 建立語音訊號之隱藏式馬可夫模型.......................23
3.3 隱藏式馬可夫模型辨識程序.............................24
四、語法規則.................................................25
4.1 文法剖析器的簡介.......................25
4.1.2文法剖析器對辨識結果的矯正 .....................25
4.2 使用HMM方法訓練語法規則模型..........................28
4.2.1 維特比演算法...................................28
五、辨識系統與CLogo的連結....................................32
六、實驗結果.................................................33
七、結論與未來展望...........................................36
7.1結論................................................36
7.2 未來展望.................................................37
參考文獻......................................................38
附錄一........................................................39
附錄二........................................................49
附錄三........................................................53

[1]PC電腦語音辨認實作，陳明熒著──台北市：旗標，民83
[2]聲霸卡之應用與語音辨識，許志興──台北市：旗標，民83
[3]Lawrence R.Rabiner, A Tutorial on Hidden Markov Models and Selected Applications in Speech Recognition ,1989
[4]Lonis Cossette.Edga Velez and Vladimir Cuperman, Speaker-Independent Isolated-Digit Recognition Based On Hidden Markov Models And Multiple Vocabulary Specific Vector Quantization ,1991
[5]Xiaolin Li and Marc Parizeau, Training Hidden Markov Models with Multiple Observations-A Combinatorial Method,2000
[6]Qiang Huo and Chorkin Chan, Contextual Vector Quantization for Speech Recognition with Discrete Hidden Markov Model,1994
[7]Jim Z.C.Lai And C.C.Lve, Fast Search Algorithms for VQ Codebook Generation,PP163-168,1996
[8]Monson H.Hayes, Statical Digital Signal Processing And Modeling
[9] L.R.Rabiner/R.W.Schafer, Digital Processing of Speech Signals
[10]Todd K.Moon And Wynn C.Stirling, Mathematical Methods and Algorithms for Signal Processing