Single Channel Speech Enhancement Technique Research Articles

Low SNR speech enhancement with DNN based phase estimation

In low Signal-to-Noise Ratio environment phase information is one of the important factor and therefore this article consider the importance of clean phase in single channel speech enhancement technique. The proposed method uses Deep Neural Network based regression model to estimate clean phase and clean amplitude for speech reconstruction. Experiments are conducted over five different noises such as factory, restaurant, car, airport and babble at different levels and result are evaluated using objective quality measures like Perceptual Evaluation of Speech Quality, Weighted Spectral Slope, Cepstrum Distance, frequency weighted segmented Signal-to-Noise Ratio and Log Likelihood Ratio. The overall quality of speech improved for factory noise by $$12\%$$ , restaurant noise by $$8\%$$ , car noise by $$13\%$$ , airport noise by $$10\%$$ and babble noise by $$14\%$$ respectively.

Subjective and Objective Analysis of Speech Enhancement Algorithms for Single Channel Speech Patterns of Indian and English Languages

ABSTRACTThis paper presents a comparative study among the seven single channel speech enhancement techniques such as spectral subtraction, Wiener filtering, minimum mean square error under speech presence uncertainty (MMSE-SPU), p-MMSE, log-MMSE and modulation channel selection (MCS). For the investigation of the capability of these techniques, 12 different practical noises on five different language databases were used. The result was analysed based on subjective and objective measure. In subjective measure SNR, peak signal-to-noise ratio (PSNR), segmental-SNR (SSNR) and mean square error (MSE) were considered, whereas for objective measure speech the intelligibility index was taken. The different language (Hindi, Kannada, Malayalam, Bengali and English) databases were taken from the Noizeus speech corpus and IIIT-H Indic speech database, while the noise database was obtained from the Noizex-92 noise corpus. The algorithms were implemented in MATLAB. The results obtained are very encouraging and helpful in the selection of single channel speech enhancement technique for practical application-based noise reduction. Further, among all the mentioned methods, MCS shows overall better performance for the five languages and 12 different practical noises.

Bayesian marginal statistics for speech enhancement using log Gabor wavelet

This paper deals with single-channel speech enhancement technique. Initially, the suitability of Log Gabor Wavelet (LGW) is investigated in speech enhancement approach and a novel speech enhancer by Bayesian Maximum a Posteriori (MAP) based Marginal Statistical Characterization (MSC) is developed. The LGW filters are traditional choice for obtaining localized frequency information and these offer the best simultaneous localization of time and frequency information. The MSC is applied in each scale of the LGW, that means a level dependent shrinkage rule is taken to suppress the background perturbations. The pdf of the LGW filtered speech coefficient is modeled with Generalized Laplacian Distribution (GLD), which allows a high approximation accuracy for Laplace distributed real and imaginary parts of the speech coefficients. The robustness of the proposed framework is tested on NOIZEUS speech corpus against seven different established speech enhancement algorithms. Experimental results show that the proposed estimator yield a higher improvement in Segmental SNR (S-SNR), lower Log Area Ratio (LAR) and Weighted Spectral Slope (WSS) distortion compared to existing speech enhancement algorithms.

A single channel speech enhancement technique exploiting human auditory masking properties

Abstract. To enhance extreme corrupted speech signals, an Improved Psychoacoustically Motivated Spectral Weighting Rule (IPMSWR) is proposed, that controls the predefined residual noise level by a time-frequency dependent parameter. Unlike conventional Psychoacoustically Motivated Spectral Weighting Rules (PMSWR), the level of the residual noise is here varied throughout the enhanced speech based on the discrimination between the regions with speech presence and speech absence by means of segmental SNR within critical bands. Controlling in such a way the level of the residual noise in the noise only region avoids the unpleasant residual noise perceived at very low SNRs. To derive the gain coefficients, the computation of the masking curve and the estimation of the corrupting noise power are required. Since the clean speech is generally not available for a single channel speech enhancement technique, the rough clean speech components needed to compute the masking curve are here obtained using advanced spectral subtraction techniques. To estimate the corrupting noise, a new technique is employed, that relies on the noise power estimation using rapid adaptation and recursive smoothing principles. The performances of the proposed approach are objectively and subjectively compared to the conventional approaches to highlight the aforementioned improvement.