论文信息 - Implementation of vocal tract length normalization for phoneme recognition on timit speech corpus

Implementation of vocal tract length normalization for phoneme recognition on timit speech corpus

Inter-speaker variability, one of the problems faced in speech recognition system, has caused the performance degradation in recognizing varied speech spoken by different speakers. Vocal Tract Length Normalization (VTLN) method is known to improve the recognition performances by compensating the speech signal using specific warping factor. Experiments are conducted using TIMIT speech corpus and Hidden Markov Model Toolkit (HTK) together with the implementation of VTLN method in order to show improvement in speaker independent phoneme recognition. The results show better recognition performance using Bigram Language Model compared to Unigram Language Model, with Phoneme Error Rate (PER) 28.8% as the best recognition performance for Bigram and PER 38.09% for Unigram. The best warp factor used for normalization in this experiment is 1.40.

Jensen Jing Lung Wong | Abdul Manan Ahmad | Mohd Shafry Mohd Rahim | Md. Sah Salam

[1] Hsiao-Wuen Hon,et al. Speaker-independent phone recognition using hidden Markov models , 1989, IEEE Trans. Acoust. Speech Signal Process..

[2] Biing-Hwang Juang,et al. Fundamentals of speech recognition , 1993, Prentice Hall signal processing series.

[3] Alfred Mertins,et al. Invariant integration features combined with speaker-adaptation methods , 2010, INTERSPEECH.

[4] Fabio Brugnara,et al. Improved automatic speech recognition through speaker normalization , 2006, Comput. Speech Lang..

[5] Li Lee,et al. A frequency warping approach to speaker normalization , 1998, IEEE Trans. Speech Audio Process..

[6] Stan Davis,et al. Comparison of Parametric Representations for Monosyllabic Word Recognition in Continuously Spoken Se , 1980 .

[7] Alfred Mertins,et al. Robust Features for Speaker-Independent Speech Recognition Based on a Certain Class of Translation-Invariant Transformations , 2009, NOLISP.

[8] Ming Liu,et al. Frequency domain correspondence for speaker normalization , 2007, INTERSPEECH.

[9] Alex Waibel,et al. Vocal Tract Length Normalization for Large Vocabulary Continuous Speech Recognition , 1997 .

[10] Li Lee,et al. Speaker normalization using efficient frequency warping procedures , 1996, 1996 IEEE International Conference on Acoustics, Speech, and Signal Processing Conference Proceedings.