Deep Residual Echo Suppression and Noise Reduction: A Multi-Input FCRN Approach in a Hybrid Speech Enhancement System.
Jan FranzenTim FingscheidtPublished in: ICASSP (2022)
Keyphrases
- noise reduction
- speech enhancement
- edge detection
- signal to noise ratio
- noisy environments
- edge preserving
- wiener filter
- single channel
- median filter
- background noise
- speech signal
- non stationary
- linear prediction
- hearing aids
- computer vision
- vocal tract
- edge information
- multi channel
- high frequency
- pattern recognition
- multiscale
- image processing