A Comparative Analysis of Channel-Robust Features and Channel Equalization Methods for Speech Recognition

The use of a speech recognition system with telephone channel environments, or different microphones, requires channel equalisation. In speech recognition, the speech models provide a bank of statistical information that can be used in the channel identification and equalisation process. The authors consider HMM-based channel equalisation, and present results demonstrating that substantial improvement can be obtained through the equalisation process. An alternative method is to use a set of features which is more robust to channel distortion. Channel distortions result in an amplitude-tilt of the speech cepstrum, and so differential cepstral features should provide a measure of immunity to channel distortions. In particular the cepstral-time feature matrix, in addition to providing a framework for representing speech dynamics, can be made robust to channel distortions. They present results demonstrating that a major advantage of cepstral-time matrices is their channel insensitive character.