TR2026-163

An S5 System based on Multi-stage Separation and Classification

    •  Saijo, K., Masuyama, Y., Boeddeker, C., Wichern, G., Richter, J., Edo, T., Le Roux, J., "An S5 System based on Multi-stage Separation and Classification", Workshop on Detection and Classification of Acoustic Scenes and Events (DCASE), October 2026.
      BibTeX TR2026-163 PDF
      • @inproceedings{Saijo2026oct,
      • author = {Saijo, Kohei and Masuyama, Yoshiki and Boeddeker, Christoph and Wichern, Gordon and Richter, Julius and Edo, Takahiro and {Le Roux}, Jonathan},
      • title = {{An S5 System based on Multi-stage Separation and Classification}},
      • booktitle = {Workshop on Detection and Classification of Acoustic Scenes and Events (DCASE)},
      • year = 2026,
      • month = oct,
      • url = {https://www.merl.com/publications/TR2026-163}
      • }
  • MERL Contacts:
  • Research Areas:

    Artificial Intelligence, Machine Learning, Speech & Audio

Abstract:

This paper describes a multi-stage separation and classi- fication system for DCASE 2026 Challenge Task 4: Spatial Semantic
Segmentation of Sound Scenes (S5). Inspired by the top-ranked system in DCASE 2025 Task 4, we adopt a cascaded framework consisting of universal sound separation (USS) with source counting, source classification, and class-aware refinement. In the first stage, a TF-
Locoformer-based USS model separates multi-channel mixtures into single-channel foreground and interference signals. Each separated signal is then classified as one of 18 foreground classes or as interference.
The separated foreground signals are further refined by another TF-
Locoformer-based model conditioned on the predicted class labels and the observed mixture. Our best system achieves a CA-PI-SDRi of 12.94 dB on the final evaluation set, ranking third overall. Notably, the system achieves a label prediction accuracy of 76.92%, the best classification performance among all submissions.