Cascaded Time + Time-Frequency Unet For Speech Enhancement: Jointly Addressing Clipping, Codec Distortions, And Gaps.
Arun Asokan NairKazuhito KoishidaPublished in: ICASSP (2021)
Keyphrases
- speech enhancement
- noisy environments
- noise reduction
- signal to noise ratio
- single channel
- speech signal
- frequency domain
- linear prediction
- vocal tract
- wiener filter
- signal processing
- multi channel
- speech recognition
- wavelet transform
- smoothing algorithm
- bitstream
- motion estimation
- additive noise
- spatial domain
- computer vision
- image sequences
- image processing