以多階層向量量化為基礎之語者辨識

Speaker Identification Based on Multistage Vector Quantization

鄭竹勝(Chu-Sheng Cheng)

指導教授 ： 謝景棠

淡江大學/工學院/電機工程學系碩士班/碩士(2007年)

https://doi.org/10.6846/TKU.2007.00755

摘要

為提高系統的抗雜訊能力，我們提出一個新的語者辨識系統。在辨識器前作ㄧ區分週期與非週期性語料的預處理，因為語音的非週期性語料易受雜訊影響且能量較弱，我們僅利用週期性語料進行語者辨認，又因屬於極零模型的線性預估倒頻譜參數(LPCC) 比梅爾刻度倒頻譜參數(MFCC)對週期性語料有較好的描述能力，我們使用LPCC作為語者辨識的特徵參數，且多階層的VQ比單階層VQ具有更好的編碼能力與需較少的容量，我們使用作為語者辨識的分類器，以提升辨識率。

關鍵字

語者辨識 ； 語音切割 ； 向量量化 ； 多階層向量量化

並列摘要

We presents an effective speaker recognition system for improving the performance in noisy environments and under various recording conditions, including microphone, common phones. In our previous works, we segment speech manually into regions of aperiodic consonant and others. As we find the characteristic of aperiodic consonant of LPCCs effect the performance of speaker identification in noisy environments. For speech feature extraction, we use LPCC and MFCC. The experimental results show that LPCC is more effective than MFCC, particularly extract form periodic corpora. For classifiers, we have tested VQ (Vector quantization) and 2-stage VQ. The experimental results show that 2-stage VQ is more effective than VQ and the 2-stage VQ is computationally more efficient than VQ.
In our experiments, to evaluate the combinations of speech features and speaker classifiers, we have used two speech corpora in this study, include TIMIT and NTIMIT database.

並列關鍵字

Speaker Identification ； Speech Segmentation ； Vector Quantization ； Multistage VQ

參考文獻

[16] W. C. Chen, C.T. Hsieh and E. Lai, “Multiband Approach to Robust Text-Independent Speaker Identification”, C.L.C.L. Process., Vol.9, NO.2, pp. 63-76, August 2004.

[14] H. C. Wang, “MAT- A project to collect Mandarin speech data through telephone networks in Taiwan”, Comput. Linguist. Chinese Language Proc., pp. 73-90, 1997.

[5] F.K. Soong, A. E. Rosenberg and B. H. Juang, “A vector quantization approach to speaker recognition”, IEEE Trans. Speech, and Signal Process., Vol. 10 , pp. 387-390, 1985.

[1] L. S. Huang and C. H. Yang, “A novel approach roust speech endpoint detection in car environments”, Proc. IEEE ICASSP-00, Vol. 3, pp. 1751-1754, June 2000.

[2] J. F. Wang, C. H. Wu, S.H. Chang, and J. Y. Lee, “A hierarchical neural network model based on a C/V segmentation algorithm for isolated Mandarin speech recognition”, IEEE Trans. Signal Process., 39, No. 9, pp. 2141-2146, 1991.