HMM-based speech enhancement applied in non-stationary noise using cepstral features and log-normal approximation

This paper proposes a hidden Markov model (HMM)-based speech enhancement method, aiming at reducing non-stationary noise from speech signals. The system is based on the assumption that the speech and the noise are additive and uncorrelated. Cepstral features are used to extract statistical information from both the speech and the noise. A priori statistical information is collected from long training sequences into ergodic hidden Markov models. Given the ergodic models for the speech and the noise a compensated model is created by means of parallel model combination, using a log-normal approximation. During compensation, the mean of every mixture in the speech and noise model is stored. The stored means are then used in the enhancement process to create the most likely speech and noise power spectral distributions using the forward algorithm combined with mixture probability. The distributions are used to generate an optimal linear Wiener filter for every observation. An evaluation of the speech enhancer working in a non-stationary noise environment is performed