Speaker-Aware Speech Enhancement with Self-Attention
- Resource Type
- Conference
- Authors
- Lin, Ju; Van Wijngaarden, Adriaan J.; Smith, Melissa C.; Wang, Kuang-Ching
- Source
- 2021 29th European Signal Processing Conference (EUSIPCO) Signal Processing Conference (EUSIPCO), 2021 29th European. :486-490 Aug, 2021
- Subject
- Bioengineering
Communication, Networking and Broadcast Technologies
Components, Circuits, Devices and Systems
Computing and Processing
Engineering Profession
Signal Processing and Analysis
Recurrent neural networks
Convolution
Green products
Europe
Speech enhancement
Data mining
Noise measurement
- Language
- ISSN
- 2076-1465
Speech enhancement aims to improve the intelligibility and quality of speech that is affected by noise. In this paper, we propose a novel speaker-aware speech enhancement (SASE) method that extracts speaker information using long short-term memory (LSTM) layers, and then uses a convolutional recurrent neural network (CRN) to embed the extracted speaker information. It is shown in a series of comprehensive experiments that only a few seconds of reference audio suffice for the proposed SASE method to perform better than LSTM and CRN baseline systems. The addition of a self-attention mechanism can further boost relevant speech-quality metrics.