Search

In this letter, we introduce a discriminative training algorithm of the basis vectors in the nonnegative matrix factorization (NMF) model for single-channel speech enhancement. The basis vectors for the clean speech and noises are estimated simultaneously during the training stage by incorporating the concept of classification from machine learning. Specifically, we consider the probabilistic generative model (PGM) of classification, which is specified by class-conditional densities, along with the NMF model. The update rules of the NMF are jointly obtained with the parameters of the class-conditional densities using the expectation–maximization (EM) algorithm, which guarantees convergence. Experimental results show that the proposed algorithm provides better performance in speech enhancement than the benchmark algorithms. [ABSTRACT FROM PUBLISHER]

• Colored-noise Kalman filter is adopted in our system, which is more component to deal with the complex noise and alleviate the speech distortion. • A multi-objective DNN is first employed to joint estimate parameters of the clean speech autoregressive (AR) model and the noise AR model. Two kinds of DNN, i.e., fully-connected feed-forward network (FNN) and long short-term memory (LSTM), are adopted. • A post subtraction technique is employed to further remove the residual noise in the Kalman-filtered speech. • The proposed system takes advantage of both the DNN based method and Kalman filtering, and has a good generalization capability in both seen and unseen noise environments. In this paper, we propose a new speech enhancement system using a deep neural network (DNN)-augmented colored-noise Kalman filter. In our system, both clean speech and noise are modelled as autoregressive (AR) processes, whose parameters comprise the linear prediction coefficients (LPCs) and the driving noise variances. The LPCs are obtained through training a multi-objective DNN that learns the mapping from the noisy acoustic features to the line spectrum frequencies (LSFs), while the driving noise variances are obtained by solving an optimization problem aiming to minimize the difference between the modelled and observed AR spectra of the noisy speech. The colored-noise Kalman filter with DNN estimated parameters is then applied to the noisy speech for denoising. Finally, a post-subtraction technique is adopted to further remove the residual noise in the Kalman-filtered speech. Extensive computer simulations show that the proposed speech enhancement system achieves significant performance gains when compared to conventional Kalman filter based algorithms as well as recent DNN-based methods under both seen and unseen noise conditions. [ABSTRACT FROM AUTHOR]

We introduce single-channel supervised speech enhancement algorithms based on regularized non-negative matrix factorization (NMF). In the proposed framework, the log-likelihood functions (LLF) of the magnitude spectra for both the clean speech and noise, based on Gaussian mixture models (GMM), are included as regularization terms in the NMF cost function. By using this proposed regularization as a priori information in the enhancement stage, we can exploit the statistical properties of both the clean speech and noise signals. For further improvement of the enhanced speech quality, we also incorporate a masking model of the human auditory system in our approach. Specifically, we construct a weighted Wiener filter (WWF) where the power spectral densities (PSD) of the speech and noise are estimated from the above mentioned NMF algorithm with the proposed regularization. The weighting factor in the WWF is selected based on a masking threshold which is obtained from the estimated PSD of the enhanced speech. Experimental results of perceptual evaluation of speech quality (PESQ), source-to-distortion ratio (SDR) and segmental signal-to-noise ratio (SNR) show that the proposed speech enhancement algorithms (i.e., regularized NMF with and without masking model) provide better performance in speech enhancement than the benchmark algorithms. [ABSTRACT FROM AUTHOR]

In this paper, we propose a new dereverberation approach based on the weighted prediction error (WPE) method implemented in the short-time Fourier transform (STFT) domain. Our main contribution is to model the temporal correlation of the STFT coefficients across analysis frames, referred to as inter-frame correlation (IFC), and exploit it in the dereverberation process. Since accurate modeling of the IFC is not tractable, we consider an approximate model wherein only a finite number of consecutive speech frames are considered correlated. It is shown that, given an estimate of the IFC matrix, the proposed approach results in a convex quadratic optimization problem with respect to the reverberation prediction weights, and a closed-form solution can be accordingly derived. Furthermore, an efficient method for the estimation of the underlying IFC matrix is developed based on the extension of a recently proposed speech variance estimator. We evaluate the performance of our approach incorporating the estimated IFC matrix and compare it to the original and several variants of the WPE method. The results reveal lower residual reverberation and higher overall quality of the enhanced speech when the proposed method is employed. [ABSTRACT FROM AUTHOR]

Most of the current approaches for speech enhancement (SE) using deep neural network (DNN) face a number of limitations: they do not exploit information contained in the phase spectrum while their high computational complexity and memory requirements make them unsuited for real-time applications. In this paper, a new phase-aware composite deep neural network (PACDNN) is introduced to address these challenges. Specifically, magnitude processing with spectral mask and phase reconstruction with phase derivative are proposed as key subtasks of the new network to simultaneously enhance the magnitude and phase spectra. Besides, the DNN is meticulously designed to take advantage of strong temporal and spectral dependencies of speech, while its components perform independently and in parallel to speed up the computation. The advantages of the proposed PACDNN model over some well-known DNN-based SE methods are demonstrated through extensive comparative experiments. • A novel phase-aware composite DNN for speech magnitude and phase enhancement. • Exploiting improved LSTM and CNN to extract a complementary set of features. • Parallel streams to speed up the computations. • Low complexity in terms of the number of trainable parameters and memory. • Extensive comparative experiments showing a competitive performance. [ABSTRACT FROM AUTHOR]

In this paper, we introduce a training and compensation algorithm of the class-conditioned basis vectors in the non-negative matrix factorization (NMF) model for single-channel speech enhancement. The main goal is to estimate the basis vectors of different signal sources in a way that prevents them from representing each other, in order to reduce the residual noise components that have features similar to the speech signal. During the proposed training stage, the basis matrices for the clean speech and noises are estimated jointly by constraining them to belong to different classes. To this end, we employ the probabilistic generative model (PGM) of classification, specified by class-conditional densities, as an a priori distribution for the basis vectors. The update rules of the NMF and the PGM parameters of classification are jointly obtained by using the variational Bayesian expectation-maximization (VBEM) algorithm, which guarantees convergence to a stationary point. Another goal of the proposed algorithm is to handle a mismatch between the characteristics of the training and test data. This is accomplished during the proposed enhancement stage, where we implement a basis compensation scheme. Specifically, we use extra free basis vectors to capture the features that are not included in the training data. Objective experimental results for different combination of speaker and noise types show that the proposed algorithm can provide better speech enhancement performance than the benchmark algorithms under various conditions. [ABSTRACT FROM AUTHOR]