Jesper Lisby Højvang

Yang Xiang, Jesper Lisby Højvang, Morten Højfeldt Rasmussen et al.

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 the 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.

Yang Xiang, Liming Shi, Jesper Lisby Højvang et al.

In this paper, we propose a novel supervised single-channel speech enhancement method combing the the Kullback-Leibler divergence-based non-negative matrix factorization (NMF) and hidden Markov model (NMF-HMM). With the application of HMM, the temporal dynamics information of speech signals can be taken into account. In the training stage, the sum of Poisson, leading to the KL divergence measure, is used as the observation model for each state of HMM. This ensures that a computationally efficient multiplicative update can be used for the parameter update of the proposed model. In the online enhancement stage, we propose a novel minimum mean-square error (MMSE) estimator for the proposed NMF-HMM. This estimator can be implemented using parallel computing, saving the time complexity. The performance of the proposed algorithm is verified by objective measures. The experimental results show that the proposed strategy achieves better speech enhancement performance than state-of-the-art speech enhancement methods. More specifically, compared with the traditional NMF-based speech enhancement methods, our proposed algorithm achieves a 5\% improvement for short-time objective intelligibility (STOI) and 0.18 improvement for perceptual evaluation of speech quality (PESQ).