Enhancement Of Noisy Speech Research Articles

An Overview of Subjective and Objective Quality Measures for Noisy Speech Enhancement Algorithms

Many signal processing algorithms have been proposed to enhance the quality and/or intelligibility of speech degraded by the additive background noise. In all speech enhancement algorithms the improvement in the quality and intelligibility are of utmost importance for ease and accuracy of information exchange. The speech quality and intelligibility can be quantified using subjective and objective measures. Subjective speech quality measures are obtained using listening tests in which human participants rate quality of the speech in accordance with a predetermined opinion scale. Objective quality measures predict perceived speech quality based on a computation of the numerical distance or distortion between the original and the processed speech. In this paper, several subjective and objective quality measures that determine the speech quality will be briefly discussed.

Nonlinear enhancement of noisy speech, using continuous attractor dynamics formed in recurrent neural networks

Here, formation of continuous attractor dynamics in a nonlinear recurrent neural network is used to achieve a nonlinear speech denoising method, in order to implement robust phoneme recognition and information retrieval. Formation of attractor dynamics in recurrent neural network is first carried out by training the clean speech subspace as the continuous attractor. Then, it is used to recognize noisy speech with both stationary and nonstationary noise. In this work, the efficiency of a nonlinear feedforward network is compared to the same one with a recurrent connection in its hidden layer. The structure and training of this recurrent connection, is designed in such a way that the network learns to denoise the signal step by step, using properties of attractors it has formed, along with phone recognition. Using these connections, the recognition accuracy is improved 21% for the stationary signal and 14% for the nonstationary one with 0db SNR, in respect to a reference model which is a feedforward neural network.

Speech Enhancement Algorithm Using Sub band Two Step Decision Directed Approach with Adaptive Weighting factor and Noise Masking Threshold

Problem statement: Speech Enhancement plays an important role in any of the speech processing systems like speech recognition, mobile communication, hearing aid. Approach: In this work, human perceptual auditory masking effect is incorporated into the single channel speech enhancement algorithm. The algorithm is based on a criterion by which the audible noise may be masked rather than being attenuated and thereby reducing the chance of distortion to speech. The basic decision directed approach is for efficient reduction of musical noise, that includes the estimation of the a priori SNR which is a crucial parameter of the spectral gain, follows the a posteriori SNR with a delay of one frame in speech frames. In this work a simple adaptive speech enhancement technique, using an adaptive sigmoid type function to determine the weighting factor of the TSDD algorithm is employed based on a sub band approach. In turn the spectral estimate is used to obtain a perceptual gain factor. Results: Objective and subjective measures like SNR, MSE, IS distance and were obtained, which shows the ability of the proposed method for efficient enhancement of noisy speech Conclusion/Recommendations: Performance assessment shows that our proposal can achieve a more significant noise reduction and a better spectral estimation of weak speech spectral components from a noisy signal as compared to the conventional speech enhancement algorithm.

Enhancement of noisy speech by temporal and spectral processing

This paper presents a noisy speech enhancement method by combining linear prediction (LP) residual weighting in the time domain and spectral processing in the frequency domain to provide better noise suppression as well as better enhancement in the speech regions. The noisy speech is initially processed by the excitation source (LP residual) based temporal processing that involves identifying and enhancing the excitation source based speech-specific features present at the gross and fine temporal levels. The gross level features are identified by estimating the following speech parameters: sum of the peaks in the discrete Fourier transform (DFT) spectrum, smoothed Hilbert envelope of the LP residual and modulation spectrum values, all from the noisy speech signal. The fine level features are identified using the knowledge of the instants of significant excitation. A weight function is derived from the gross and fine weight functions to obtain the temporally processed speech signal. The temporally processed speech is further subjected to spectral domain processing. Spectral processing involves estimation and removal of degrading components, and also identification and enhancement of speech-specific spectral components. The proposed method is evaluated using different objective and subjective quality measures. The quality measures show that the proposed combined temporal and spectral processing method provides better enhancement, compared to either temporal or spectral processing alone.

Noise suppression using non-linear speech model

Noise suppression has been a long-standing problem in digital speech processing. For average signal-to-noise ratio (SNR) and stationary noise there are commercially available software products or equipment. Therefore, the subject of the recent research is the enhancement of noisy speech in case of non-stationary noise. The noise suppression algorithm, presented in this paper, is based on dimension-embedding, and separates speech from noise in the transformed domain. The algorithm has been tested on several noisy versions of a Hungarian sentence using three different noise-types and seven different SNRs. The gains achieved correspond to the published data.

Processing Noisy Speech for Enhancement

The main objective of this paper is to provide an overview of the major techniques that have been proposed for enhancement of noisy speech. This paper also proposes a method for enhancement of noisy speech. The proposed method is motivated as an attempt to minimize the limitations of conventional spectral subtraction method for enhancement of noisy speech. The proposed method involves three steps. In the first step speech and nonspeech regions are detected from the degraded speech signal. In the second step, for each speech region, noise components are estimated from the preceding noise regions and are subtracted by conventional spectral subtraction method. In the third step, speech components are enhanced further from the spectral subtracted speech signal for reducing the musical noise, which is the artifact of spectral subtraction method. The processed speech signals from the proposed method seem to be better perceptually compared to that of the spectral subtraction method.

Dual ${H}_{\infty}$ Algorithms for Signal Processing— Application to Speech Enhancement

This paper deals with the joint signal and parameter estimation for linear state-space models. An efficient solution to this problem can be obtained by using a recursive instrumental variable technique based on two dual Kalman filters. In that case, the driving process and the observation noise in the state-space representation for each filter must be white with known variances. These conditions, however, are too strong to be always satisfied in real cases. To relax them, we propose a new approach based on two dual Hinfin filters. Once a new observation of the disturbed signal is available, the first Hinfin algorithm uses the latest estimated parameters to estimate the signal, while the second Hinfin algorithm uses the estimated signal to update the parameters. In addition, as the Hinfin filter behavior depends on the choice of various weights, we present a way to recursively tune them. This approach is then studied in the following cases: (1) consistent estimation of the AR parameters from noisy observations and (2) speech enhancement, where no a priori model of the additive noise is required for the proposed approach. In each case, a comparative study with existing methods is carried out to analyze the relevance of our solution.

Improved Global Soft Decision Using Smoothed Global Likelihood Ratio for Speech Enhancement

In this letter, we propose an improved global soft decision for noisy speech enhancement. From an investigation of statistical model-based speech enhancement, it is discovered that a global soft decision has a fundamental drawback at the speech tail regions of speech signals. For that reason, we propose a new solution based on a smoothed likelihood ratio for the global soft decision. Performances of the proposed method are evaluated by subjective tests under various environments and show better results compared with the our previous work.

Kalman tracking of linear predictor and harmonic noise models for noisy speech enhancement

This paper presents a speech enhancement method based on the tracking and denoising of the formants of a linear prediction (LP) model of the spectral envelope of speech and the parameters of a harmonic noise model (HNM) of its excitation. The main advantages of tracking and denoising the prominent energy contours of speech are the efficient use of the spectral and temporal structures of successive speech frames and a mitigation of processing artefact known as the ‘musical noise’ or ‘musical tones’. The formant-tracking linear prediction (FTLP) model estimation consists of three stages: (a) speech pre-cleaning based on a spectral amplitude estimation, (b) formant-tracking across successive speech frames using the Viterbi method, and (c) Kalman filtering of the formant trajectories across successive speech frames. The HNM parameters for the excitation signal comprise; voiced/unvoiced decision, the fundamental frequency, the harmonics’ amplitudes and the variance of the noise component of excitation. A frequency-domain pitch extraction method is proposed that searches for the peak signal to noise ratios (SNRs) at the harmonics. For each speech frame several pitch candidates are calculated. An estimate of the pitch trajectory across successive frames is obtained using a Viterbi decoder. The trajectories of the noisy excitation harmonics across successive speech frames are modeled and denoised using Kalman filters. The proposed method is used to deconstruct noisy speech, de-noise its model parameters and then reconstitute speech from its cleaned parts. Experimental evaluations show the performance gains of the formant tracking, pitch extraction and noise reduction stages.

Complex laplacian probability density function for noisy speech enhancement

In this letter, we propose a new approach to speech enhancement based on a complex Laplacian probability density function (pdf). With a use of a goodness-of-fit (GOF) test, we discover that the complex Laplacian pdf is more desirable to describe noisy speech distribution than the conventional Gaussian pdf for speech enhancement. The likelihood ratio (LR) is computed and then applied to computation of the speech absence probability in the speech enhancement algorithm. Our evaluation shows that the proposed complex Laplacian pdf scheme provides further speech quality enhancement compared with the conventional method through a number of objective tests.

Multiple statistical models for soft decision in noisy speech enhancement

Most speech enhancement algorithms are based on the assumption that speech and noise are both Gaussian in the discrete cosine transform (DCT) domain. For further enhancement of noisy speech in the DCT domain, we consider multiple statistical distributions (i.e., Gaussian, Laplacian and Gamma) as a set of candidates to model the noise and speech. We first use the goodness-of-fit (GOF) test in order to measure how far the assumed model deviate from the actual distribution for each DCT component of noisy speech. Our evaluations illustrate that the best candidate is assigned to each frequency bin depending on the Signal-to-Noise-Ratio (SNR) and the Power Spectral Flatness Measure (PSFM). In particular, since the PSFM exhibits a strong relation with the best statistical fit we employ a simple recursive estimation of the PSFM in the model selection. The proposed speech enhancement algorithm employs a soft estimate of the speech absence probability (SAP) separately for each frequency bin according to the selected distribution. Both objective and subjective tests are performed for the evaluation of the proposed algorithms on a large speech database, for various SNR values and types of background noise. Our evaluations show that the proposed soft decision scheme based on multiple statistical modeling or the PSFM provides further speech quality enhancement compared with recent methods through a number of subjective and objective tests.

Inter-frame modeling of DFT trajectories of speech and noise for speech enhancement using Kalman filters

In this paper a time–frequency estimator for enhancement of noisy speech signals in the DFT domain is introduced. This estimator is based on modeling the time-varying correlation of the temporal trajectories of the short-time (ST) DFT components of the noisy speech signal using autoregressive (AR) models. The time-varying trajectory of the DFT components of speech in each channel is modeled by a low-order AR process incorporated in the state equation of Kalman filters. The parameters of the Kalman filters are estimated recursively from the estimates of the signal and noise in DFT channels. The issue of convergence of the Kalman filters’ statistics during the noise-only periods is addressed. A method is incorporated for restarting of Kalman filters, after long periods of noise-dominated activity in a DFT channel, to mitigate distortions of the onsets of speech activity. The performance of the proposed method with and without AR modeling of the DFT trajectories of noise for the enhancement of noisy speech is evaluated and compared with the MMSE log-amplitude speech estimator, parametric spectral subtraction and Wiener filter. Evaluation results show that the incorporation of spectral–temporal information through Kalman filters results in reduced residual noise and improved perceived quality of speech.

Warped discrete cosine transform-based noisy speech enhancement

In this paper, a warped discrete cosine transform (WDCT)-based approach to enhance the degraded speech under background noise environments is proposed. For developing an effective expression of the frequency characteristics of the input speech, the variable frequency warping filter is applied to the conventional discrete cosine transform (DCT). The frequency warping control parameter is adjusted according to the analysis of spectral distribution in each frame. For a more accurate analysis of spectral characteristics, the split-band approach in which the global soft decision for speech presence is performed in each band separately is employed. A number of subjective and objective tests show that the WDCT-based enhancement method yields better performance than the conventional DCT-based algorithm.

Time-Varying Noise Estimation for Speech Enhancement and Recognition Using Sequential Monte Carlo Method

We present a method for sequentially estimating time-varying noise parameters. Noise parameters are sequences of time-varying mean vectors representing the noise power in the log-spectral domain. The proposed sequential Monte Carlo method generates a set of particles in compliance with the prior distribution given by clean speech models. The noise parameters in this model evolve according to random walk functions and the model uses extended Kalman filters to update the weight of each particle as a function of observed noisy speech signals, speech model parameters, and the evolved noise parameters in each particle. Finally, the updated noise parameter is obtained by means of minimum mean square error (MMSE) estimation on these particles. For efficient computations, the residual resampling and Metropolis-Hastings smoothing are used. The proposed sequential estimation method is applied to noisy speech recognition and speech enhancement under strongly time-varying noise conditions. In both scenarios, this method outperforms some alternative methods.

Cepstrum-domain acoustic feature compensation based on decomposition of speech and noise for asr in noisy environments

This paper presents a set of acoustic feature pre-processing techniques that are applied to improving automatic speech recognition (ASR) performance on noisy speech recognition tasks. The principal contribution of this paper is an approach for cepstrum-domain feature compensation in ASR which is motivated by techniques for decomposing speech and noise that were originally developed for noisy speech enhancement. This approach is applied in combination with other feature compensation algorithms to compensating ASR features obtained from a mel-filterbank cepstrum coefficient front-end. Performance comparisons are made with respect to the application of the minimum mean squared error log spectral amplitude (MMSE-LSA) estimator based speech enhancement algorithm prior to feature analysis. An experimental study is presented where the feature compensation approaches described in the paper are found to greatly reduce ASR word error rate compared to uncompensated features under environmental and channel mismatched conditions.

Performance results of noisy speech enhancement using undecimated wavelets and spectral peak enhancement

This paper presents a new speech enhancement algorithm based upon an undecimated discrete wavelet transform followed by spectral peak enhancement. Experimental data is presented which demonstrates improved automatic speech recognition scores in noisy speech. The new algorithm first removes noise by soft-thresholding in the undecimated wavelet domain. However, the noise reduction process leaves residual noise that impairs intelligibility. In the new algorithm, residual noise removal after the undecimated wavelet denoising scheme is achieved by a nonlinear modification of the Fourier spectrum to enhance the spectral peaks relative to the valleys. Using 30 sentences at each SNR, evaluation of the new algorithm was conducted with the Connected Speech Test (CST), a procedure that ensures that the speech sentences have equal difficulty. For additive white Gaussian noise, significant improvements are observed in the experiment results and it is shown that an effective speech enhancement is obtained by spectral peak enhancement with small decomposition level. The results were variable with babble noise and office noise. Performance results will be presented

Array signal processing applicable to hearing aids

For the use of hearing aids in noisy environments, noise reduction or speech enhancement techniques are being investigated by many researchers. In this report, array signal-processing techniques and systems applicable to hearing aids, which are being developed by the authors, are introduced. A delay-and-sum beamformer, in which the microphone-array outputs are summed with appropriate delays and weights, can focus on a signal coming from a certain direction and can reduce environmental noise. Moreover, by scanning the beam by controlling the delays and the weights, the directions of the sound sources can also be known. By combining these techniques, a noise reduction system, which tracks the movement of a talker and suppresses the environmental noise, can be constructed. In this report, hardware equipped with two DSPs (TMS320C40), which realizes the above method, is introduced. Moreover, by extending the beamformer to stereo outputs, a listener can perceive the direction from which the sound is coming. In this report, simulation results of the stereo beamformer are also briefly described.

Frequency domain adaptive postfiltering for enhancement of noisy speech

This paper presents a new frequency-domain approach to implement an adaptive postfilter for enhancement of noisy speech. The postfilter is described by a set of DFT coefficients which suppress noise in the spectral valleys and allow for more noise in formant regions which is masked by the speech signal. First, we perform an LPC analysis of the noisy speech and calculate the log magnitude spectrum of the input speech. After identifying the formants and valleys (by a new method), the log magnitude spectrum is modified to obtain the postfilter coefficients. The filtering operation is also done in the frequency domain through an FFT and an overlap-add strategy to get the postfiltered speech. Experimental results on 8-kHz-sampled speech show that this new frequency-domain approach results in enhanced speech of better perceptual quality than obtained by a time-domain method. This new method is especially efficient in eliminating high frequency noise and in preserving the weaker, high frequency formants in sonorant sounds.

Comparison of noisy speech enhancement algorithms in terms of LPC perturbation

A number of noisy speech enhancement algorithms are experimentally compared in terms of linear predictive coding (LPC) perturbations. The enhancement algorithms considered are simple spectral subtraction, spectral oversubtraction with use of a spectral floor, spectral subtraction with residual noise removal, and time-domain and frequency-domain adaptive minimum mean-square-error filtering. LPC perturbations considered are LPC cepstral distance, log likelihood ratios, and weighted likelihood ratio.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Coherence‐based enhancement of noisy speech

A standard coherence measure has been used to give an estimate of signal‐to‐noise ratio (certainty indicators) as a function of frequency. The two spectra which are used for this measure are the noisy speech spectrum and a model spectrum derived from pitch period estimates of the noisy speech. It is intended that the certainty indicators be used as either weights for matches in automatic recognition or as multiplicative weights for reconstruction of speech with improved intelligibility. Two cases of noisy speech have been studied: single words contaminated by bandpass‐filtered white noise (S1) and sentences contaminated by unfiltered white noise (S2). For case S1 it has been found that the certainty indicators will pinpoint the noisy band for voiced speech and will, as expected, not reliably indicate the noisy band for unvoiced speech. For case S2, preliminary results indicate that the quality of the enhanced speech has dropped, yet the intelligibility is slightly improved. [Work supported by NSF and Boeing Computer Services.]