Robust Speech Recognition for Unknown Communication Channels Pont-à-Mousson, France April 17-18, 1997

Channel Normalisation Using Phase-Corrected RASTA

Johan de Veth, Louis Boves

A2RT, Department of Language and Speech, University of Nijmegen, Nijmegen, The Netherlands

Recently, we proposed an extension to the classical RASTA technique. The new method consists of classical RASTA filtering followed by a phase correction operation. In this manner, the influence of the communication channel is as effectively removed as with classical RASTA. However, our proposal does not introduce a left-context dependency like classical RASTA. Therefore the new method is better suited for automatic speech recognition based on context-independent modeling with Gaussian mixture hidden Markov models. In this paper we introduce an implementation of phase-corrected RASTA suited for real-time processing. In the context of connected digit recognition over the phone using context-independent phone-based models, we show that word error rate for this implementation is more than 20% lower compared to a real-time implementation of the cepstrum mean subtraction channel normalisation method.

Full Paper

Bibliographic reference.  Veth, Johan de / Boves, Louis (1997): "Channel normalisation using phase-corrected RASTA", In RSR-1997, 119-122.