Dilated U-net based approach for multichannel speech enhancement from First-Order Ambisonics recordings.
Amélie BoscaAlexandre GuérinLauréline PerotinSrdan KiticPublished in: EUSIPCO (2020)
Keyphrases
- single channel
- speech enhancement
- multi channel
- independent component analysis
- frequency domain
- prior information
- higher order
- signal processing
- linear prediction
- wiener filter
- wavelet decomposition
- speech signal
- sound source
- non stationary
- high frequency
- multiscale
- noisy environments
- brain activity
- pattern recognition