Incorporating voice onset time to improve letter recognition accuracies

Author

Niyogi, Partha ; Ramesh, Padma

Author_Institution

Lucent Technol., AT&T Bell Labs., Murray Hill, NJ, USA

Volume

1

fYear

1998

fDate

12-15 May 1998

Firstpage

13

Abstract

We consider the possibility of incorporating distinctive features into a statistically based speech recognizer. We develop a two pass strategy for recognition with a standard HMM based first pass followed by a second pass that performs an alternative analysis to extract class-specific features. For the voiced/voiceless distinction on stops for an alphabet recognition task, we show that a linguistically motivated acoustic feature exists (the VOT), provides superior separability to standard spectral measures, and can be automatically extracted from the signal to reduce error rates by 48.7% over state of the art HMM systems

Keywords

error correction; feature extraction; hidden Markov models; linguistics; speech recognition; statistical analysis; HMM systems; alphabet recognition; class-specific feature extraction; distinctive features; error correcting device; error rate reduction; letter recognition accuracies; linguistically motivated acoustic feature; second pass; spectral measures; standard HMM based first pass; statistically based speech recognizer; stops; two pass strategy; voice onset time; voiced/voiceless distinction; Acoustic measurements; Automatic speech recognition; Error analysis; Error correction; Feature extraction; Hidden Markov models; Measurement standards; Performance analysis; Speech recognition; Standards development;

fLanguage

English

Publisher

ieee

Conference_Titel

Acoustics, Speech and Signal Processing, 1998. Proceedings of the 1998 IEEE International Conference on

Conference_Location

Seattle, WA

ISSN

1520-6149

Print_ISBN

0-7803-4428-6

Type

conf

DOI

10.1109/ICASSP.1998.674355

Filename

674355