Title :
Normalizing the vocal tract length for speaker independent speech recognition
Author :
Lin, Qiguang ; Che, ChiWei
Author_Institution :
Center for Comput. Aids for Ind. Productivity, Rutgers Univ., Piscataway, NJ, USA
Abstract :
We overall vocal-tract length differs between individuals. The difference is more apparent between males and females. In this work, a new method is presented which normalizes the overall vocal-tract length in the cepstrum domain. By properly selecting the upper frequency of a FFT spectrum, the derived cepstrum coefficients exhibit a pattern which is less susceptible to the length variations. The proposed method is evaluated in isolated-word speech recognition experiments where a recognizer was exclusively trained on male speech and tested on female speech, and vice versa. The method elevated the word recognition accuracy from 57.0 to 78.8% and from 70.5 to 86.1%, respectively.<>
Keywords :
cepstral analysis; fast Fourier transforms; speech recognition; FFT spectrum; cepstrum coefficients; cepstrum domain; female speech; isolated-word speech recognition experiments; length variations; male speech; speaker independent speech recognition; upper frequency; vocal tract length normalization; word recognition accuracy; Cepstrum; Frequency; Hidden Markov models; Pharynx; Speech analysis; Speech recognition; Speech synthesis; Testing; Transfer functions;
Journal_Title :
Signal Processing Letters, IEEE