TPARN: Triple-Path Attentive Recurrent Network for Time-Domain Multichannel Speech Enhancement.
Ashutosh PandeyBuye XuAnurag KumarJacob DonleyPaul CalamiaDeLiang WangPublished in: ICASSP (2022)
Keyphrases
- speech enhancement
- single channel
- recurrent networks
- frequency domain
- multi channel
- linear prediction
- recurrent neural networks
- feed forward
- signal to noise ratio
- noise reduction
- noisy environments
- biologically inspired
- neural network
- wiener filter
- speech signal
- independent component analysis
- spatial domain
- visual attention
- lossless compression
- vocal tract
- prior information
- wavelet decomposition
- artificial neural networks
- denoising
- smoothing algorithm
- wavelet transform
- edge preserving
- sound source
- eye movements
- focus of attention
- filter bank
- image coding