Abstract
Compact time-frequency separators that mask the mixture and refine through a shared cell face two limits. First, a bounded multiplicative mask only scales a mixture bin, so where overlapping components cancel, the estimate stays small. Second, a shared cell applies the same weights to every time-frequency token at every step, so enlarging it adds compute everywhere. We present SEAL (Sparse Expert routing with Additive Latent reconstruction) to address both. For reconstruction, a zero-sum additive residual bounded by the local mixture amplitude lets estimates be nonzero where components cancel yet still sum to the mixture. For routing, a query built from acoustic and inter-step evidence sends each token to one of six residual experts, and a norm cap keeps the step cue from overriding clear acoustic evidence. On EchoSet, SEAL (small) surpasses TIGER (small) by 0.31 dB SI-SDRi with 28% fewer parameters and 2.9 times fewer MACs, and SEAL (large) is within 0.07 dB SI-SDRi of TIGER (large) at 3.1 times fewer MACs.
Keywords
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Hu, S. C., Lin, Z. X., Hung, J. W., & Lee, H. S. (2026). SEAL: Mixture-Closed Additive Reconstruction and Refinement-Aware Expert Routing for Efficient Speech Separation. https://omanscience.com/en/articles/seal-mixture-closed-additive-reconstruction-and-refinement-aware-expert-routing-for-efficient-speech-separation
MLA 9
Hu, Shao-Chun, et al. "SEAL: Mixture-Closed Additive Reconstruction and Refinement-Aware Expert Routing for Efficient Speech Separation." https://omanscience.com/en/articles/seal-mixture-closed-additive-reconstruction-and-refinement-aware-expert-routing-for-efficient-speech-separation.
Chicago (author–date)
Hu, Shao-Chun, Zi-Xiang Lin, Jeih-Weih Hung, and Hung-Shin Lee. 2026. "SEAL: Mixture-Closed Additive Reconstruction and Refinement-Aware Expert Routing for Efficient Speech Separation." https://omanscience.com/en/articles/seal-mixture-closed-additive-reconstruction-and-refinement-aware-expert-routing-for-efficient-speech-separation.
Harvard
Hu, S. C., Lin, Z. X., Hung, J. W. and Lee, H. S. (2026) 'SEAL: Mixture-Closed Additive Reconstruction and Refinement-Aware Expert Routing for Efficient Speech Separation', Available at: https://omanscience.com/en/articles/seal-mixture-closed-additive-reconstruction-and-refinement-aware-expert-routing-for-efficient-speech-separation.
Vancouver
Hu SC, Lin ZX, Hung JW, Lee HS. SEAL: Mixture-Closed Additive Reconstruction and Refinement-Aware Expert Routing for Efficient Speech Separation. https://omanscience.com/en/articles/seal-mixture-closed-additive-reconstruction-and-refinement-aware-expert-routing-for-efficient-speech-separation
IEEE
S. C. Hu, Z. X. Lin, J. W. Hung, and H. S. Lee, "SEAL: Mixture-Closed Additive Reconstruction and Refinement-Aware Expert Routing for Efficient Speech Separation," https://omanscience.com/en/articles/seal-mixture-closed-additive-reconstruction-and-refinement-aware-expert-routing-for-efficient-speech-separation.