Supervised and Unsupervised Speech Enhancement Using Nonnegative Matrix Factorization

By Nasser Mohammadiha, Paris Smaragdis, Arne Leijon. IEEE Transactions on Audio, Speech, and Language Processing, 2013

398 citations recorded by the source corpus.

Reducing the interference noise in a monaural noisy speech signal has been a challenging task for many years. Compared to traditional unsupervised speech enhancement methods, e.g., Wiener filtering, supervised approaches, such as algorithms based on hidden Markov models (HMM), lead to higher-quality enhanced speech signals. However, the main practical difficulty of these approaches is that for each noise type a model is required to be trained a priori. In this paper, we investigate a new class of supervised speech denoising algorithms using nonnegative matrix factorization (NMF). We propose a novel speech enhancement method that is based on a Bayesian formulation of NMF (BNMF). To circumvent the mismatch problem between the training and testing stages, we propose two solutions. First, we use an HMM in combination with BNMF (BNMF-HMM) to derive a minimum mean square error (MMSE) estimator for the speech signal with no information about the underlying noise type. Second, we suggest a scheme to learn the required noise BNMF model online, which is then used to develop an unsupervised speech enhancement system. Extensive experiments are carried out to investigate the performance of the proposed methods under different conditions. Moreover, we compare the performance of the developed algorithms with state-of-the-art speech enhancement schemes using various objective measures. Our simulations show that the proposed BNMF-based methods outperform the competing algorithms substantially.

Subfields: Artificial Intelligence, Computational Mechanics, Signal Processing

View Paper: https://www.semanticscholar.org/paper/31113b1ab61b41b2b03e7b34c377c48f94de9aff

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Jinyu Li
2. Dong Yu
3. Li Deng
4. Junichi Yamagishi
5. Yifan Gong
6. Haizhou Li
7. Chin-Hui Lee
8. Zhuo Chen

In Computational Mechanics

1. Junichi Yamagishi
2. Chin-Hui Lee
3. Marc Delcroix
4. Jinyu Li
5. Jun Du
6. Yong Xu
7. Lirong Dai
8. Zhuo Chen

In Signal Processing

1. Jinyu Li
2. Dong Yu
3. Junichi Yamagishi
4. Shinji Watanabe
5. Yifan Gong
6. Chin-Hui Lee
7. Marc Delcroix
8. Li Deng

Citations in the corpus

• Interactive Speech and Noise Modeling for Speech Enhancement (2020, methodology): “…signal processing and modeling based methods mostly take the other way (see Figure 1 (b)), i.e. estimating noise or building noise models for speech enhancement (Boll 1979; Hendriks, Heusdens, and Jensen 2010; Wang and Brookes 2017; Wilson et al. 2008; Mohammadiha, Smaragdis, and Leijon 2013).”
• An Overview of Noise-Robust Automatic Speech Recognition (2014, methodology): “This is a source separation approach, and non-negative matrix factorization (NMF) [166] has been shown to be a very successful method [167], [168], and can directly benefit noise-robust ASR [163]– [165], [169].”
• An Overview of Voice Conversion and Its Challenges: From Statistical Modeling to Deep Learning (2020, methodology): “Non-negative matrix factorization (NMF) [139] is an effective data mining technique that has been widely used, especially for reconstruction of high quality signals, such as in speech enhancement [140], [141], speech de-noising [142], [143], noise and speech estimation [144].”
• Text-informed speech enhancement with deep neural networks (2015, methodology): “In the last decade, a great progress has been made in the field of speech enhancement research by introducing statistical speech models, such as Non-negative Matrix Factorization (NMF)-based model [2–4], Gaussian mixture model (GMM) [5] and hiddenMarkov model (HMM) [6].”
• Cracking the cocktail party problem by multi-beam deep attractor network (2017, methodology): “erties of speech, such as pitch continuity, harmonic structures, common onsets etc., and this can be achieved by using statistical models [10], rule-based models [3, 11], or decompositionbased models [12, 13, 14, 15]. In multi-channel systems, the separation process can exploit the spatial propertie”
• Multi-objective learning and mask-based post-processing for deep neural network based speech enhancement (2017, background): “Non-negative matrix factorization (NMF) based speech enhancement [7, 8] was one notable example in which speech and noise basis models were learned separately from training speech and noise databases.”
• TCNN: Temporal Convolutional Neural Network for Real-time Speech Enhancement in the Time Domain (2019, neutral): “Traditional speech enhancement approaches include spectral subtraction methods [1], Wiener filtering [2], statistical model-based methods [3] and nonnegative matrix factorization [4].”