Unifying Speech Enhancement and Separation with Gradient Modulation for End-to-End Noise-Robust Speech Separation.
Yuchen HuChen ChenHeqing ZouXionghu ZhongEng Siong ChngPublished in: CoRR (2023)
Keyphrases
- end to end
- speech enhancement
- noisy environments
- single channel
- noise reduction
- signal to noise ratio
- speech signal
- speech recognition
- noisy speech
- background noise
- multi channel
- sound source
- linear prediction
- multipath
- text localization and recognition
- smoothing algorithm
- additive noise
- vocal tract
- automatic speech recognition
- edge preserving
- edge detection
- hidden markov models
- prior information
- wavelet transform