Personalized speech enhancement combining band-split RNN and speaker attentive module.
Xiaohuai LeZhongshu HouLi ChenChao HeYiqing GuoCheng ChenXianjun XiaJing LuPublished in: CoRR (2023)
Keyphrases
- speech enhancement
- vocal tract
- noisy environments
- recurrent neural networks
- speaker verification
- noise reduction
- speech signal
- signal to noise ratio
- single channel
- speech recognition
- audio visual
- automatic speech recognition
- visual attention
- background noise
- speaker identification
- neural network
- wiener filter
- optical flow
- multiscale