We proposed an augmented cepstral mean normalization

algorithm that differentiates noise and speech during

normalization, and computes a different mean for each. The new

procedure reduced the error rate slightly for the case of sameenvironment

testing, and significantly reduced the error rate by

25% when an environmental mismatch exists over the case of

standard cepstral mean normalization.

PDF file

In  Proc. of the IEEE Workshop on Automatic Speech Recognition

Details