An hmm-based cepstral-domain speech enhancement system
C.W. Seymour, Mahesan Niranjan · 1994
This paper describes a method of enhancing speech corrupted by additive uncorrelated noise. The approach adopted is to use cepstral-domain hidden Markov models to determine statistics of the clean speech and noise processes. A compensated model of speech corrupted by noise is generated using parallel model combination. MMSE and linear nonhomogeneous estimators of the clean speech signal are derived. The enhancement system gives natural -sounding speech without the artifacts introduced by systems such as spectral subtraction. HMM recognition tests performed on the enhanced speech using the NOISEX-92 database show a significant reduction in error rate. 1. INTRODUCTION This paper is concerned with the problem of enhancing speech corrupted by additive uncorrelated noise. The uses of enhancement are: a) to construct a signal which, according to some criteria, sounds better than the noisy speech; and b) to estimate some parameters of the clean speech as a front end to a computer speech reco...