Noise Tracking Using DFT Domain Subspace Decompositions

Abstract

Abstract-All discrete Fourier transform (DFT) domain-based speech enhancement gain functions rely on knowledge of the noise power spectral density (PSD). Since the noise PSD is unknown in advance, estimation from the noisy speech signal is necessary. An overestimation of the noise PSD will lead to a loss in speech quality, while an underestimation will lead to an unnecessary high level of residual noise. We present a novel approach for noise tracking, which updates the noise PSD for each DFT coefficient in the presence of both speech and noise. This method is based on the eigenvalue decomposition of correlation matrices that are constructed from time series of noisy DFT coefficients. The presented method is very well capable of tracking gradually changing noise types. In comparison to state-of-the-art noise tracking algorithms the proposed method reduces the estimation error between the estimated and the true noise PSD. In combination with an enhancement system the proposed method improves the segmental SNR with several decibels for gradually changing noise types. Listening experiments show that the proposed system is preferred over the state-of-the-art noise tracking algorithm.Index Terms-Discrete Fourier transform (DFT) domain subspace decompositions, noise tracking, speech enhancement.

Unbiased MMSE-Based Noise Power Estimation With Low Complexity and Low Tracking Delay

“…As in [19], we compare the estimated noise power σ 2 N to the reference σ 2 N in terms of the log-error distortion measure.…”

“…Some examples are the discrete Fourier transform (DFT)-subspace approach [19], or minimum mean-square error (MMSE)-based approaches [20] [21]. Although DFTsubspace-based approaches lead to quite some improvement for non-stationary noise sources compared to, e.g., MS-based spectral noise power estimators [22], computationally they are rather demanding.…”

Environment dependent noise tracking for speech enhancement

“…Rangachari and Loizou (2006), proposed advancements over the MCRA scheme that adapts faster to changing noise levels. This approach was further extended by Hendriks et al (2008) where they performed minima tracking on an eigen decomposition subspace instead of the FFT bins. Examples of approaches that use decompositions other than FFT include Chatlan and Soraghan (2009).…”

A Two-Step Noise Estimation Algorithm for Noisy Speech Enhancement

“…However, this noise PSD estimation is sensitive to its outliers and its variance is about twice as large as the variance of conventional noise estimation. Moreover, this method may occasionally attenuate low energy phonemes, particularly if the minimum search window is too short [6]. Soft decision (SD), the other well known noise power estimation technique adapts the noise statistics based on the uncertainty of speech absence [7].…”