CEPSTRAL PARAMETER COMPENSATION FOR HMM RECOGNITION IN NOISE

This paper describes a method of adapting a continuous density HMM rec ogniser trained on clean cepstral speech data to make it robust to noi se. The technique is based on parallel model combination (PMC) in whic h the parameters of corresponding pairs of speech and noise states are combined to yield a set of compensated parameters. It improves on ear lier cepstral mean compensation methods in that it also adapts the var iances and as a result can deal with much lower SNRs. The PMC method i s evaluated on the NOISEX-92 noise database and shown to work well dow n to 0 dB SNR and below for both stationary and non-stationary noises. Furthermore, for relatively constant noise conditions, there is no ad ditional computational cost at run-time.