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

Abstract

Abstract-Recently, it has been proposed to estimate the noise power spectral density by means of minimum mean-square error (MMSE) optimal estimation. We show that the resulting estimator can be interpreted as a voice activity detector (VAD)-based noise power estimator, where the noise power is updated only when speech absence is signaled, compensated with a required bias compensation. We show that the bias compensation is unnecessary when we replace the VAD by a soft speech presence probability (SPP) with fixed priors. Choosing fixed priors also has the benefit of decoupling the noise power estimator from subsequent steps in a speech enhancement framework, such as the estimation of the speech power and the estimation of the clean speech. We show that the proposed SPP approach maintains the quick noise tracking performance of the bias compensated MMSE-based approach while exhibiting less overestimation of the spectral noise power and an even lower computational complexity.Index Terms-Noise power estimation, speech enhancement.

Speech enhancement using modulation-domain Kalman filtering with active speech level normalized log-spectrum global priors

“…Random segments of noise from the noise signals are used [29]. The external noise estimation is based on [30] [29]. For pre-cleaning in Fig.…”

A Novel Approach of Speech Enhancement by Implementing Cellular Automata Algorithm

“…For this method it is assume that noise signal is stationary or slowly changing with time. Minimum mean square error (MMSE) estimators [7] and its updated versions [8], these methods do not require any prior knowledge about the speech and noise signals, nor any training stage beforehand, so they are highly flexible and allow implementation in various situations. However, these algorithms usually assume that the noise is stationary and are thus not good at dealing with nonstationary noise types, especially under low signal-to-noise (SNR) conditions.…”

Online Monaural Speech Enhancement Based on Periodicity Analysis and A Priori SNR Estimation

“…It can be seen that the estimated noise levels well follow the sudden changes of the true noise levels. Please note that here the estimated noise levels are compared with the true noise levels, not with the recursive filtering smoothed levels of the true noise as in [4]. …”

“…A review of the progress of approach can be found in chapter 9 of [1] and chapter 6 in [3]. The algorithm described in [4] is among the best of this class of algorithms. However, all of the algorithms based on the statistical framework have to assume that the spectrum levels of the noise change slowly frame by frame to make them distinguished from the spectrum levels of the speech which change fast.…”