This article focuses on leveraging deep representation learning (DRL) for speech enhancement (SE). In general, the performance of the deep neural network (DNN) is heavily dependent on the learning of data representation. However, the DRL's importance is often ignored in many DNN-based SE algorithms. To obtain a higher quality enhanced speech, we propose a two-stage DRL-based SE method through adversarial training. In the first stage, we disentangle different latent variables because disentangled representations can help DNN generate a better enhanced speech. Specifically, we use the$\beta$-variational autoencoder (VAE) algorithm to obtain the speech and noise posterior estimations and related representations from the observed signal. However, since the posteriors and representations are intractable and we can only apply a conditional assumption to estimate them, it is difficult to ensure that these estimations are always pretty accurate, which may potentially degrade the final accuracy of the signal estimation. To further improve the quality of enhanced speech, in the second stage, we introduce adversarial training to reduce the effect of the inaccurate posterior towards signal reconstruction and improve the signal estimation accuracy, making our algorithm more robust for the potentially inaccurate posterior estimations. As a result, better SE performance can be achieved. The experimental results indicate that the proposed strategy can help similar DNN-based SE algorithms achieve higher short-time objective intelligibility (STOI), perceptual evaluation of speech quality (PESQ), and scale-invariant signal-to-distortion ratio (SI-SDR) scores. Moreover, the proposed algorithm can also outperform recent competitive SE algorithms.

Recently, variational autoencoder (VAE), a deep representation learning (DRL) model, has been used to perform speech enhancement (SE). However, to the best of our knowledge, current VAE-based SE methods only apply VAE to model speech signal, while noise is modeled using the traditional non-negative matrix factorization (NMF) model. One of the most important reasons for using NMF is that these VAE-based methods cannot disentangle the speech and noise latent variables from the observed signal. Based on Bayesian theory, this paper derives a novel variational lower bound for VAE, which ensures that VAE can be trained in supervision, and can disentangle speech and noise latent variables from the observed signal. This means that the proposed method can apply the VAE to model both speech and noise signals, which is totally different from the previous VAE-based SE works. More specifically, the proposed DRL method can learn to impose speech and noise signal priors to different sets of latent variables for SE. The experimental results show that the proposed method can not only disentangle speech and noise latent variables from the observed signal, but also obtain a higher scale-invariant signal-to-distortion ratio and speech quality score than the similar deep neural network-based (DNN) SE method.

In this paper, we propose a novel non-negative matrix factorization (NMF) and hidden Markov model (NMF-HMM) based speech enhancement algorithm, which employs a Poisson mixture model (PMM). Compared to the previously proposed NMF-HMM method, the new algorithm, termed PMM-NMF-HMM, uses the Poisson mixture distribution for the state conditional likelihood function for a HMM rather than the single Poisson distribution. This means that there are the more basis matrices that can be used to model the speech and noise signals, so more signal information can be captured by the resulting model. The proposed method is supervised and thus includes a training and an enhancement stage. It is shown that, in the training stage, the proposed method can be implemented efficiently using multiplicative update (MU) for the model parameters, much like the NMF-HMM algorithm. In the speech enhancement stage, which can be performed online, a novel PMM-NMF-HMM minimum mean-square error (MMSE) estimator is developed. The experimental results indicate that the PMM-NMF-HMM method can obtain higher short-time objective intelligibility (STOI) and perceptual evaluation of speech quality (PESQ) score than NMF-HMM. Additionally, the method also outperforms other state-of-the-art NMF- based supervised speech enhancement algorithms.

In this paper, we present a novel supervised Non-negative Matrix Factorization (NMF) speech enhancement method, which is based on Hidden Markov Model (HMM) and Kullback- Leibler (KL) divergence (NMF-HMM). Our algorithm applies theHMMto capture the timing information, so the temporal dynamics of speech signal can be considered by comparing with the traditional NMF-based speech enhancement method. More specifically, the sum of Poisson, leading to the KL divergence measure, is used as the observation model for each state of HMM. This ensures that the parameter update rule of the proposed algorithm is identical to the multiplicative update rule, which is quick and efficient. In the training stage, this update rule is applied to train the NMF-HMM model. In the online enhancement stage, a novel minimum mean-square error (MMSE) estimator that combines the NMF-HMM is proposed to conduct speech enhancement. The performance of the proposed algorithm is evaluated by perceptual evaluation of speech quality (PESQ) and short-timeobjective intelligibility (STOI). The experimental results indicate that the STOI score of proposed strategy is able to outperform 7% than current state-of-the-art NMF-based speech enhancement methods.

Recently, deep neural networks (DNNs) have become the mainstream strategy for speech enhancement task because it can achieve the higher speech quality and intelligibility than the traditional methods. However, these DNN-based methods always need a large number of parallel corpus consisting of clean speech and noise to produce noisy data for the training of the DNN in order to improve the generalization of the network. As a result, this implies that many noisy speech signals that are collected in real environment cannot be used to train the DNN because of the lack of corresponding clean speech and noise. Additionally, as we know, noise varies with the time and scenario, so we cannot obtain parallel speech and noise due to infinite noise data and some limited speech data. Thus, the network training with unparallel speech and noise data is essential for the generalization of the network. To address this problem, we propose a novel parallel-data-free speech enhancement method, in which the cycle-consistent generative adversarial network (CycleGAN) and multi-objective learning are employed. Our method is also able to make best use of the benefits of multi-objective learning. On the training stage, we utilize two different encoders to encode the features of clean speech and noisy speech, respectively. Then, two forward generators are immediately used to predict the ideal time-frequency (T-F) mask and log-power spectrum (LPS) of clean speech. Two inverse generators are applied to map the magnitude spectrum (MS) and LPS of noisy speech, respectively. In addition, four discriminators are used to distinguish the real speech features from the generated features. Two encoders, four generators and four discriminators are simultaneously trained by using adversarial, identity-mapping, latent similarity and cycle-consistent loss. On the test stage, we directly utilize the forward generators and encoders to acquire the enhanced speech. The experimental results indicate that the proposed approach is able to achieve the better speech enhancement performance than the reference methods. Moreover, the proposed method is also effective to improve speech quality and intelligibility when the networks are trained under the parallel data.