TPARN: Triple-path Attentive Recurrent Network for Time-domain Multichannel Speech Enhancement.
Ashutosh PandeyBuye XuAnurag KumarJacob DonleyPaul CalamiaDeLiang WangPublished in: CoRR (2021)
Keyphrases
- speech enhancement
- single channel
- recurrent networks
- frequency domain
- linear prediction
- multi channel
- recurrent neural networks
- noise reduction
- feed forward
- biologically inspired
- noisy environments
- independent component analysis
- signal to noise ratio
- spatial domain
- neural network
- prior information
- wiener filter
- sound source
- image coding
- vocal tract
- probabilistic model
- information retrieval
- learning algorithm
- feature extraction
- optical flow
- speech signal
- subband
- denoising
- signal processing
- filter bank