Real-Time Neural Speech Enhancement Based on Temporal Refinement Network and Channel-Wise Gating Methods
Neural speech enhancement systems have seen dramatic improvements in performance recently. However, it is still difficult to create systems that can operate in real-time, with low delay, low complexity, and causality. In this paper, we propose a temporal and channel attention framework for a U-Net-based speech enhancement architecture that uses short analysis frame lengths. Specifically, we propose an attention-based temporal refinement network (TRN) that estimates convolutional features subject to the importance of temporal location. By adding the TRN output to the channel-attentive convolution output, we can further enhance speech-related features even in low-attentive channel outputs. To further improve the representation power of the convolutional features, we also apply a squeeze-and-excitation (SE)-based channel attention mechanism for three different network modules: main convolutional blocks after processing the TRN, skip connections, and residual connections in the bottleneck recurrent neural network (RNN) layer. In particular, a channel-wise gate architecture placed on the skip connections and residual connections reliably controls the data flow, which avoids transferring redundant information to the following stages. We show the effectiveness of the proposed TRN and channel-wise gating methods by visualizing the spectral characteristics of the corresponding features, evaluating overall enhancement performance, and performing ablation studies in various configurations. Our proposed real-time enhancement system outperforms several recent neural enhancement models in terms of quality, model size, and complexity.
|6||International Journal||Zainab Alhakeem, Se-In Jang, Hong-Goo Kang "Disentangled Representations in Local-Global Contexts for Arabic Dialect Identification" in Transactions on Audio, Speech, and Language Processing, 2024|
|5||International Journal||Hyungchan Yoon, Changhwan Kim, Seyun Um, Hyun-Wook Yoon, Hong-Goo Kang "SC-CNN: Effective Speaker Conditioning Method for Zero-Shot Multi-Speaker Text-to-Speech Systems" in IEEE Signal Processing Letters, vol.30, pp.593-597, 2023|
|4||International Journal||Jinyoung Lee, Hong-Goo Kang "Real-Time Neural Speech Enhancement Based on Temporal Refinement Network and Channel-Wise Gating Methods" in Digital Signal Processing, vol.133, 2023|
|3||International Journal||Taemin Kim, Yejee Shin, Kyowon Kang, Kiho Kim, Gwanho Kim, Yunsu Byeon, Hwayeon Kim, Yuyan Gao, Jeong Ryong Lee, Geonhui Son, Taeseong Kim, Yohan Jun, Jihyun Kim, Jinyoung Lee, Seyun Um, Yoohwan Kwon, Byung Gwan Son, Myeongki Cho, Mingyu Sang, Jongwoon Shin, Kyubeen Kim, Jungmin Suh, Heekyeong Choi, Seokjun Hong, Huanyu Cheng, Hong-Goo Kang, Dosik Hwang & Ki Jun Yu "Ultrathin crystalline-silicon-based strain gauges with deep learning algorithms for silent speech interfaces" in Nature Communications, vol.13, 2022|
|2||International Journal||Jinyoung Lee, Hong-Goo Kang "Two-Stage Refinement of Magnitude and Complex Spectra for Real-Time Speech Enhancement" in IEEE Signal Processing Letters, vol.29, pp.2188-2192, 2022|
|1||International Journal||Kyungguen Byun, Seyun Um, Hong-Goo Kang "Length-Normalized Representation Learning for Speech Signals" in IEEE Access, vol.10, pp.60362-60372, 2022|