Complex-Valued Spatial Autoencoders for Multichannel Speech Enhancement.
Mhd Modar HalimehWalter KellermannPublished in: ICASSP (2022)
Keyphrases
- complex valued
- speech enhancement
- single channel
- real valued
- linear prediction
- multi channel
- recurrent neural networks
- noise reduction
- denoising
- linear equations
- noisy environments
- signal to noise ratio
- wiener filter
- activation function
- multi valued
- frequency domain
- neural network
- independent component analysis
- vocal tract
- prior information
- speech signal
- hidden markov models
- feature selection