Speech Enhancement Network with Unsupervised Attention using Invariant Information Clustering

• Published in: 2021 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) pp. 406 - 409
• Main Authors: Sugiura, Yosuke, Nagamori, Shunta, Shimamura, Tetsuya
• Format: Conference Proceeding
• Language: English
• Published: APSIPA 14.12.2021
• Subjects: Information processing; Noise measurement; Speech enhancement; Training

In this paper, we propose a new framework for speech enhancement using supervised attention trained by Invariant Information Clustering (IIC). For suppressing an overfitting in the speech enhancement network, the multitask learning with the speaker-invariant information is adopted at the latent repr...