ISCA Tutorial and Research Workshop on Statistical and Perceptual Audition (SAPA2006)

Pittsburgh, PA, USA
September 16, 2006

Improving the Noise-Robustness of Mel-Frequency Cepstral Coefficients for Speech Processing

Sourabh Ravindran (1), David V. Anderson (1), Malcolm Slaney (2)

(1) School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, GA, USA
(2) Yahoo! Research, Sunnyvale, CA, USA

In this paper we study the noise-robustness of mel-frequency cepstral coefficients (MFCCs) and explore ways to improve their performance in noisy conditions. Improvements based on a more accurate model of the early auditory system are suggested to make the MFCC features more robust to noise while preserving their class discrimination ability. Speech versus non-speech classification and speech recognition are chosen to evaluate the performance gains afforded by the modifications.

Full Paper

Bibliographic reference.  Ravindran, Sourabh / Anderson, David V. / Slaney, Malcolm (2006): "Improving the noise-robustness of mel-frequency cepstral coefficients for speech processing", In SAPA-2006, 48-52.