MossFormer2: Combining Transformer and RNN-Free Recurrent Network for Enhanced Time-Domain Monaural Speech Separation

Benchmark Model Rank Results
speech-separation-on-libri2mixMossFormer2 (w speed perturb)#1SI-SDRi: 22.2
speech-separation-on-libri2mixMossFormer2 (w/o DM)#3SI-SDRi: 21.7
speech-separation-on-whamMossFormer2#2SI-SDRi: 18.1
speech-separation-on-whamrMossFormer2#4SI-SDRi: 17.0
speech-separation-on-wsj0-2mixMossFormer2 (L)#5SI-SDRi: 24.1Number of parameters (M): 55.7
speech-separation-on-wsj0-3mixMossFormer2#1SI-SDRi: 22.2