Multi-channel speech enhancement with ad-hoc sensors has been a challenging task. Speech model guided beamforming algorithms are able to recover natural sounding speech, but the speech models tend to be oversimplified or the inference would otherwise be too complicated. On the other hand, deep learning based enhancement approaches are able to learn complicated speech distributions and perform e...

Research has shown that degrading acoustic background noise innuences speech quality across phoneme classes in a non-uniform manner. This results in variable quality performance of many speech enhancement algorithms in noisy environments. A phoneme classiication procedure is proposed which directs single-channel constrained speech enhancement. The procedure performs broad phoneme class partitio...

This paper discusses the problem of single-channel speech enhancement in variable noise-level environment. Commonly used, singlechannel subtractive-type speech enhancement algorithms always assume that the background noise level is fixed or slowly varying. In fact, the background noise level may vary quickly. This condition usually results in wrong speech/noise detection and wrong speech enhanc...

This paper addresses the study of the speech intelligibility enhancement. The speech model, noise sources, perceptual aspects of speech, and performance evaluation are reviewed. The intelligibility enhancement system based on spectral subtraction technique is investigated. Spectral density estimation device based on the algorithm of smoothed periodograms is analysed. Determination of the silenc...

We give an overview of speech quality measures and apply them to the typical shortcomings of speech enhancement algorithms. This is important for comparing different approaches to noise reduction, and it suggests ways to improve a speech enhancement algorithm. Objective quality measures are compared to the subjective quality evaluation of human listeners by means of extensive listening tests. A...

A method of speech enhancement is developed that reconstructs clean speech from a set of acoustic features using a sinusoidal model of speech. This is a significant departure from traditional filtering-based methods of speech enhancement. A major challenge with this approach is to estimate accurately the acoustic features (voicing, fundamental frequency, spectral envelope) from noisy speech. Th...

The enhancement of speech degraded by nonstationary noises and low signal-to-noise ratio (SNR) conditions is a high demanding but challenging task. We present a robust and effective single channel speech enhancement algorithm under the framework of Nonnegative Matrix Factorization (NMF). Considering the sparse property of speech and low-rank property of nonstationary noise, a new formulation fo...