TPARN: Triple-path attentive recurrent network for time-domain multichannel speech enhancement

Multichannel speech enhancement model using triple-path attentive recurrent network

BibTex: