MMSE speech enhancement based on GMM and solving an over-determined system of equations

Abstract

A new and effective algorithm is proposed in this paper based on Gaussian Mixture Modelling (GMM) and Minimum Mean Square Error (MMSE) criterion for speech enhancement where no assumption is made on the nature or stationarity of the noise. No Voice Activity Detection (VAD) or any other means is used to estimate the input Signal to Noise Ratio (SNR). The mean vectors of the mixture models of spectral magnitudes derived from models of speech and different noise sources power spectra are used to form sets of over-determined system of equations, as many as noise source candidates, whose solutions lead to the MMSE estimations of speech and additive noise spectral magnitudes. The corresponding power spectra are then used for noise suppression by applying Wiener filtering carried out on overlapping frames. The input SNR is estimated and the nature of the noise involved is determined as by-products of the method used. Results are compared with codebook constrained methods that have shown very good results but suffer from long processing times. It is shown that, at the cost of a slight lower improvement in SNR and PESQ score, the new algorithm reduces the computation time to one fifth which makes it suitable for practical applications. (Abstract)

GMM based multi-stage Wiener filtering for low SNR speech enhancement

“…We compare our approach with Chehresa et al [6]. Figure 1 and 2 show the speech enhancement performance over the two noise types.…”

“…The number of Gaussians for the speech and noise GMMs are chosen to be K s = 6 and K v = 9 following a trial and error approach proposed in [5], [6]. In order to obtain better denoising performance at low SNR levels, we use β = 2 and γ = 1.…”

“…We will show that multi-stage speech enhancement can improve performance by providing more accurate estimates of the noise PSD. Simulations show that a better speech enhancement is achieved compared to the traditional Wiener filtering method [6].…”

“…Single-channel speech enhancement methods can be classified mainly into statistical-based approaches and datadriven approaches. Statistical-based approaches include spectral subtraction [2], Wiener filtering [3][4][5][6][7], the minimum mean-square error (MMSE) short-time spectral amplitude estimation [8], etc. In general, these methods tend to perform poorly in non-stationary noisy conditions and thus often assume stationarity to obtain good performance.…”

“…In our contribution, an extension of the Gaussian mixture model (GMM) based single-channel speech enhancement approach in [6] for modeling noise power spectral densities (PSDs) is presented. We improve the performance through a multi-stage parametric Wiener filter at low SNR levels and non-stationary noise conditions.…”