臺灣博碩士論文加值系統

為了減少調適機制所需的訓練語料，本論文研究的語者調適方法，將焦點放於MFCC特徵參數之層次。先為各個語者建立個人的座標系統，用以分解MFCC向量得到座標係數，再依此係數作座標系統間的對映，而獲得語者調適的效果。為了驗證所研究方法的可行性，我們分別作了單母音、多母音、含鼻音音節及雙字詞的辨識實驗，不特定語者方式的錯誤率原先分別是30.33%、20.72%、38.25%及21.22%，而語者調適方式的辨識錯誤率則分別減少成3.32%、9.78%、22.52%及12.27%。

In this thesis, a speaker adaptation method is developed. This method needs only a small quantity of training utterances because the adaptation mechanism is operated on the level of MFCC feature parameter. First, an individual coordinate system is built for each new speaker in order that his MFCC feature vectors can be decomposed into coordinate coefficients of the system. Then, the coordinate coefficients are directly mapped as coefficients of the coordinate system for a target person. Even though this mechanism is simple, it can indeed obtain good adaptation performance. To verify the performance of our adaptation method, we have executed several recognition experiments under different conditions. The conditions are for different kinds of vocabularies, including sing-vowel vocabulary, multi-vowel vocabulary, nasal-containing syllable vocabulary and dissyllabic word vocabulary. In speaker non-adapted mode, the original recognition error rates are 30.3%, 20.7%, 38.3% and 21.3% respectively. However, in speaker adapted mode, the error rates are reduced to 3.3%, 9.8%, 22.5% and 12.3% respectively.

摘要 I
ABSTRACT II
誌 謝 III
圖表索引 VI
第一章 緒論 1
1.1 研究動機及目的 1
1.2 語者調適研究之回顧 2
1.2.1 語者正規化 3
1.2.1.1 頻率軸之頻率校正 3
1.2.1.2 聲道長度正規化 5
1.2.2 模型參數的調適 6
1.2.2.1 貝氏調適 6
1.2.2.2 最大相似線性回歸 8
1.2.3 基於代表性空間之調適 9
1.2.3.1 代表性語音 9
1.2.3.2 代表性語音最大似然度線性回歸 11
1.2.3.3 代表性語音最大後機率線性回歸 12
1.3 本論文之研究方法 13
1.4 論文架構 15
第二章 語者調適方法與處理步驟 16
2.1 本論文研究之調適方法 16
2.2 語者調適之實驗步驟 18
2.3 特徵參數求取之步驟 21
2.4 基底向量訂定與座標系統建立 23
2.4.1 基底向量選取 24
2.4.2 基底向量正規化 24
2.4.3 正交化之前的嘗試 25
2.4.4 正交化基底向量 30
2.5 求取座標係數 32
2.6 反求特徵向量 33
2.7 樣式比對 34
第三章 單母音之辨識實驗 36
3.1 母音發音分段之比較 36
3.2 母音發音遍數之比較 39
3.3 單母音辨識實驗 41
第四章 多母音之辨識實驗 43
4.1 多母音之辨識實驗 43
4.2 含鼻音音節之辨識實驗 46
第五章 雙字詞之辨識實驗 50
5.1 雙字詞之辨識實驗 50
5.2 無聲子音部分之處理 51
5.3 CMS之處理 56
5.4 線上即時辨識 57
5.4.1 處理流程 57
5.4.2 操作步驟 61
第六章 結論 65
參考文獻 67

[1] T. Lee, F. Zheng, W. Wu, “Reference point alignment frequency warp method for speaker adaptation”, WCCC-ICSP 2000, vol. 2, pp. 756 -759, 2000.
[2] T. Claes, I. Dologlou, L. ten Bosch, D. van Compernolle, “A novel feature transformation for vocal tract length normalization in automatic speech recognition”, IEEE Trans. on Speech and Audio Processing, vol. 6, no. 6, pp. 549 -557, November, 1998.
[3] 謝宏坤, 語音說明中搜尋任意定義之關鍵詞的研究, 國立台灣科技大學電機工程學研究所碩士論文, 2000.
[4] L. Rabiner and B.-H. Juang, Fundamentals of Speech Recognition, Prentice Hall, Englewood Cliffs, 1993.
[5] C.-H. Lee, C.-H. Lin, B.-H. Juang, “A study on speaker adaptation of the parameters of continuous density hidden Markov models”, IEEE Trans. on Signal Processing, Vol. 39, pp. 806 -814, April, 1991.
[6] Q. Huo, C. Chan, C. H. Lee, “Bayesian adaptive learning of the parameters of hidden Markov model for speech recognition”, IEEE Trans. on Speech and Audio Processing, Vol. 3, pp. 334 -345, September, 1995.
[7] 廖文偉, 以向量空間為基礎之語者調適技術, 國立臺灣大學電信工程學研究所碩士論文, 2000.
[8] C.J. Leggetter and P.C. Woodland, “Maximum Likelihood Linear Regression for Speaker Adaptation of Continuous Density Hidden Markov Models”, Computer Speech and Language, Vol. 9, pp. 171-185, 1995.
[9] 李上銘, 語音辨認中基於主成份分析之進一步技術, 國立臺灣大學電信工程學研究所碩士論文, 2001.
[10] R. Kuhn, P. Nguyen, J.-C. Junqua, L. Goldwasser, N. Niedzielski, S. Fincke, K. Field and M. Contolini, “Eigenvoices for speaker adaptation”, Proc. ICSLP, pp.1771-1774, 1998.
[11] K.-T. Chen, W.-W. Liau, H.-M. Wang, and L.-S. Lee, "Fast Speaker Adaptation using Eigenspace-based Maximum Likelihood Linear Regression", in Proc. International Conference on Spoken Language Processing (ICSLP2000), Beijing, Oct., 2000.
[12] N.J.-C. Wang, S.S.-M. Lee, F. Seide, L.-S. Lee, “Rapid speaker adaptation using a priori knowledge by eigenspace analysis of MLLR parameters”, IEEE Proc. ICASSP, Vol. 1, pp. 345-348, 2001.
[13] K.-T. Chen, H.-M. Wang, “Eigenspace-based maximum a posteriori linear regression for rapid speaker adaptation”, IEEE Proc. ICASSP, Vol. 1, pp. 317-320, 2001.
[14] O. Douglas, Speech Communication: human and machine 2nd. Ed., IEEE press, New York, 2000.
[15] R.O. Duda and P. E. Hart, Pattern Classification and Scene Analysis, Wiley, New York, 1973.
[16] S. J. Leon, Linear Algebra with applications, 5th. Ed., Prentice Hall, 1998.