Non-Linear Speech Processing (NOLISP 03)

May 20-23, 2003
Le Croisic, France

Audio Source Separation with One Sensor for Robust Speech Recognition

L. Benaroya, Frédéric Bimbot, G. Gravier, R. Gribonval

IRISA (INRIA & CNRS), Equipe METISS, Rennes, France

In this paper, we address the problem of noise compensation in speech signals for robust speech recognition. Several classical denoising methods in the field of speech and signal processing are compared on speech corrupted by music, which correspond to a frequent situation in broadcast news transcription tasks. We also present two new source separation techniques, namely adaptiveWiener filtering and adaptive shrinkage. These techniques rely on the use of a dictionary of spectral shapes to deal with the non stationarity of the signals. The algorithms are first compared on the source separation task and assessed in terms of average distortion. Their effect on the entire transcription system is eventually compared in terms of word error rate. Results show that the proposed adaptive Wiener filter approach yields a significant improvement of the transcription accuracy at signal/noise ratios greater than 15 dB.

Full Paper

Bibliographic reference.  Benaroya, L. / Bimbot, Frédéric / Gravier, G. / Gribonval, R. (2003): "Audio source separation with one sensor for robust speech recognition", In NOLISP-2003, paper 030.