臺灣博碩士論文加值系統

語音辨識目前最熱門的方法就是隱藏式馬可夫模型。在近幾年甚至已被拿來應用在語音評分上。本篇論文提出一個階層化演算法，用以改良隱藏式馬可夫模型用在音素語音辨識的正確率和效率，藉於音素分群機制的加入，將辨識過程分為兩個階段，第一階段判斷該語音訊號所在的音素族群，第二階段判別屬於該族群中的那個音素。在這樣階層化的方法下，降低辨識的計算量進而達到辨識效率上的改進。進而應用此改良之音素語音辨識技術實作出一套英語口語教學系統，根據精熟學習理論來幫助學生練習英語單字或句子發音。本英語口語教學系統同時具備了發音錯誤分析弁遄A針對發音、語調、節奏、音量四個向度上來告訴學生應該改進的部份，讓學生可以藉此改善其英語口語能力。

One of the most popular methods for modern speech recognition is Hidden Markov Models (HMMs), which has also been in use in speech evaluation. This thesis is aimed to propose an algorithm with improved recognition rate and efficiency for HMMs used in phoneme recognition. By adding the mechanism of phoneme clustering we can divide the recognition process into two steps: 1) determining the phoneme cluster where the speech signal belongs to, and 2) determining the phoneme which belongs to the cluster. The recognition efficiency can be improved since the amount of computation is decreased as a result of this gradual strategy. A teaching system of spoken English is subsequently devised based on the related techniques. The system can help learners practice pronunciation of English words and sentences, according to the Mastery Learning theory. Meanwhile, the teaching system of spoken English analyzes pronunciation errors and advises learners on the four dimensions they can strengthen, i.e. pronunciation, intonation, rhythm, and volume. In this way, learners’ spoken English can be improved.

Chapter 1 Introduction 1
1.1 Overview 1
1.2 Motivation 2
1.3 Organization of this Thesis 3
Chapter 2 Background 4
2.1 Speech Signal Pre-Processing 4
2.1.1 Frame Blocking 5
2.1.2 Endpoint Detection 6
2.2 Feature Extraction 9
2.3 Hidden Markov Models (HMMs) 12
2.4 Viterbi Algorithm 13
2.5 K-Means Clustering Algorithm 14
2.6 Speech Evaluation Method 15
2.7 Mastery Learning 15
Chapter 3 Algorithm 19
3.1 Speech Recognition Based on HMMs 19
3.2 Hierarchical HMM Match Algorithm 20
3.2.1 Phoneme Clustering Method 22
3.2.2 Phoneme Clustering Adaptive Method 22
3.2.3 Confusing Phoneme 24
3.3 Algorithm Analysis 25
Chapter 4 System Architecture 27
4.1 Teaching System of Spoken English 27
4.2 Requirements Gathering 29
4.3 Design 30
4.4 Implementation 36
Chapter 5 Experimental Results 43
5.1 The Speech Database 43
5.2 Phonemes Clustering Results 43
5.3 Hierarchical HMM Recognition Results 45
Chapter 6 Conclusions and Future Works 47
6.1 Conclusions 47
6.2 Future Works 47
References 49

[1]C. Myers, L.R. Rabiner, and A. E. Rosenberg, “Performance tradeoff in Dynamic Time Warping algorithms for isolated word recognition,” IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. ASSP-28, December 1980.
[2]H. Sakoe and S. Chiba, “Dynamic Programming Optimization for Spoken Word Recognition,” IEEE Transactions on ASSP, vol.26, pp. 43-49, February 1978.
[3]J. C. Junqua and H. Hermansky, “Evaluation and optimization of perceptually-based ASR front-end,” IEEE Transactions on Speech, and Audio Processing, vol. 1, pp. 39-48, January 1993.
[4]J. H. Block, Ed., “Mastery Learning: theory and practice,” New York: Holt, Rinehart and Winston, 1971.
[5]J. Makhoul, “Spectral analysis of speech by linear prediction,” IEEE Transactions on Audio and Electro acoustics, vol. 21, pp. 140-148, June 1973.
[6]L. Henson, T. Dews, M. Lotto, J. Tetzlaff, E. Dannefer, “A Mastery Learning model for assessing competency of medical students using portfolios,” Journal of Clinical Anesthesia, vol. 17, pp. 663-664, December 2005.
[7]L. R. Rabiner and B. H. Juang, “An introduction to Hidden Markov Models,” IEEE ASSP MAGAZINE, 1986.
[8]P. Angkititrakul and J. H. L. Hanson, “Advances in phone-based modeling for automatic accent classification,” IEEE Transactions on Speech and Audio Processing, vol. 14, pp. 634-646, March 2006.
[9]S. B. Davis and P. Mermelstein, “Comparison of parametric representations for monosyllabic word recognition in continuously spoken sentences,” IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. 28, pp. 357-366, August 1980.
[10]Y. Konig and N. Morgan, “Supervised and unsupervised clustering of the speaker space for connectionist speech recognition,” IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. 1, pp. 545-548, April 1993.
[11]J. S. Jang of National Tsing Hua University, “Signal processing and recognition, “ July 2006, http://neural.cs.nthu.edu.tw/jang/books/audioSignalProcessing/.
[12]羅瑞麟，以語音辨識與評分輔助口說英文學習，清華大學資訊系統與應用研究所，2003.