Sub-convolutional U-Net with transformer attention network for end-to-end single-channel speech enhancement.
Sivaramakrishna YechuriSunny Dayal VanambathinaPublished in: EURASIP J. Audio Speech Music. Process. (2024)
Keyphrases
- end to end
- single channel
- speech enhancement
- multi channel
- noise reduction
- noisy environments
- independent component analysis
- frequency domain
- wiener filter
- signal to noise ratio
- prior information
- speech signal
- wavelet decomposition
- signal processing
- vocal tract
- smoothing algorithm
- linear prediction
- multiscale
- visual attention
- multiresolution
- automatic speech recognition
- spatial domain
- background noise
- wavelet transform