Papers › Deformable Temporal Convolutional Networks for Monaural Noisy Reverberant Speech Separation

Deformable Temporal Convolutional Networks for Monaural Noisy Reverberant Speech Separation

27 Oct 2022arXiv:2210.15305archive 2025-07-28

William Ravenscroft, Stefan Goetze, Thomas Hain

Speech separation models are used for isolating individual speakers in many speech processing applications. Deep learning models have been shown to lead to state-of-the-art (SOTA) results on a number of speech separation benchmarks. One such class of models known as temporal convolutional networks (TCNs) has shown promising results for speech separation tasks. A limitation of these models is that they have a fixed receptive field (RF). Recent research in speech dereverberation has shown that the optimal RF of a TCN varies with the reverberation characteristics of the speech signal. In this work deformable convolution is proposed as a solution to allow TCN models to have dynamic RFs that can adapt to various reverberation times for reverberant speech separation. The proposed models are capable of achieving an 11.1 dB average scale-invariant signalto-distortion ratio (SISDR) improvement over the input signal on the WHAMR benchmark. A relatively small deformable TCN model of 1.3M parameters is proposed which gives comparable separation performance to larger and more computationally complex models.

PaperPDFCode

Code

jwr1995/dtcn officialmentioned in papermentioned on GitHubpytorch report
jwr1995/pubsep mentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Speech DereverberationSpeech Separation

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Speech Separation WHAMR! Deformable TCN + Dynamic Mixing MACs (G) 3.7 #15 of 18 Archive leaderboard report
Speech Separation WHAMR! Deformable TCN + Dynamic Mixing Number of parameters (M) 3.6 #15 of 18 Archive leaderboard report
Speech Separation WHAMR! Deformable TCN + Dynamic Mixing SDRi 10.3 #15 of 18 Archive leaderboard report
Speech Separation WHAMR! Deformable TCN + Dynamic Mixing SI-SDRi 11.1 #15 of 18 Archive leaderboard report
Speech Separation WHAMR! Deformable TCN + Shared Weights + Dynamic Mixing MACs (G) 3.7 #17 of 18 Archive leaderboard report
Speech Separation WHAMR! Deformable TCN + Shared Weights + Dynamic Mixing Number of parameters (M) 1.3 #17 of 18 Archive leaderboard report
Speech Separation WHAMR! Deformable TCN + Shared Weights + Dynamic Mixing SDRi 9.5 #17 of 18 Archive leaderboard report
Speech Separation WHAMR! Deformable TCN + Shared Weights + Dynamic Mixing SI-SDRi 10.1 #17 of 18 Archive leaderboard report
Speech Separation WSJ0-2mix Deformable TCN + Dynamic Mixing MACs (G) 3.7 #32 of 40 Archive leaderboard report
Speech Separation WSJ0-2mix Deformable TCN + Dynamic Mixing Number of parameters (M) 3.6 #32 of 40 Archive leaderboard report
Speech Separation WSJ0-2mix Deformable TCN + Dynamic Mixing SDRi 17.4 #32 of 40 Archive leaderboard report
Speech Separation WSJ0-2mix Deformable TCN + Dynamic Mixing SI-SDRi 17.2 #32 of 40 Archive leaderboard report
Speech Separation WSJ0-2mix Deformable TCN + Shared Weights + Dynamic Mixing MACs (G) 3.7 #34 of 40 Archive leaderboard report
Speech Separation WSJ0-2mix Deformable TCN + Shared Weights + Dynamic Mixing Number of parameters (M) 1.3 #34 of 40 Archive leaderboard report
Speech Separation WSJ0-2mix Deformable TCN + Shared Weights + Dynamic Mixing SDRi 16.3 #34 of 40 Archive leaderboard report
Speech Separation WSJ0-2mix Deformable TCN + Shared Weights + Dynamic Mixing SI-SDRi 16.1 #34 of 40 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

ConvolutionDeformable Convolution

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections