Switching Variational Auto-Encoders for Noise-Agnostic Audio-Visual Speech Enhancement.
Mostafa SadeghiXavier Alameda-PinedaPublished in: ICASSP (2021)
Keyphrases
- audio visual
- speech enhancement
- noisy environments
- noise reduction
- signal to noise ratio
- multi modal
- single channel
- speech signal
- multi stream
- background noise
- visual data
- wiener filter
- vocal tract
- visual information
- linear prediction
- multi channel
- sound source
- additive noise
- multimedia
- speech recognition
- edge detection
- audio features
- smoothing algorithm
- image segmentation
- non stationary
- image coding
- wavelet transform
- optical flow
- multiscale