Low-complexity artificial noise suppression methods for deep learning-based speech enhancement algorithms

Abstract

Deep learning-based speech enhancement algorithms have shown their powerful ability in removing both stationary and non-stationary noise components from noisy speech observations. But they often introduce artificial residual noise, especially when the training target does not contain the phase information, e.g., ideal ratio mask, or the clean speech magnitude and its variations. It is well-known that once the power of the residual noise components exceeds the noise masking threshold of the human auditory system, the perceptual speech quality may degrade. One intuitive way is to further suppress the residual noise components by a postprocessing scheme. However, the highly non-stationary nature of this kind of residual noise makes the noise power spectral density (PSD) estimation a challenging problem. To solve this problem, the paper proposes three strategies to estimate the noise PSD frame by frame, and then the residual noise can be removed effectively by applying a gain function based on the decision-directed approach. The objective measurement results show that the proposed postfiltering strategies outperform the conventional postfilter in terms of segmental signal-to-noise ratio (SNR) as well as speech quality improvement. Moreover, the AB subjective listening test shows that the preference percentages of the proposed strategies are over 60%.

CNN-Based Identification of Parkinson’s Disease from Continuous Speech in Noisy Environments

“…An objective evaluation of the Wiener filter noise suppression performance was performed in this work by using the signal-to-noise ratio (SNR) and signal-to-noise ratio improvement (SNRI) as speech enhancement measures, and the mean square error (MSE) as signal fidelity measure [52][53][54]. Each is defined as follows.…”

DMF-Net: A decoupling-style multi-band fusion model for full-band speech enhancement

“…This paper proposes a decoupling-style multiband fusion model, dubbed DMF-Net for full-band SE step by step. To be specific, we adopt a two-step strategy which consists of a pre-trained complex-domain-based SE network (LF-Net) for the low-frequency band (0-8 kHz) and two magnitude-based SE networks (MF-Net and HF-Net) for the middle-and high-frequency bands (8)(9)(10)(11)(12)(13)(14)(15)(16). In LF-Net, inspired by the preliminary study [9], we employ a decoupling-style strategy for wide-band speech enhancement by three sub-networks.…”

“…Note that the estimated low-frequency features are also fed into MF-Net and HF-Net to provide extra guidance. Finally, we fuse the low-, middle-and high-frequency regions to obtain the full-band speech, which is then fed into a low-complexity postprocessing module to further suppress the residual noise [12].…”