On Temporal Guidance and Iterative Refinement in Audio Source Separation
This work solves the problem of more accurate sound source separation for audio processing applications, but it is incremental as it builds on existing methods with iterative refinements.
The paper tackles the problem of spatial semantic segmentation of sound scenes by addressing the lack of fine-grained temporal information in conventional two-stage pipelines, resulting in significant improvements in audio tagging and source separation performance, as evidenced by a second-place finish in the DCASE Challenge 2025 Task 4.
Spatial semantic segmentation of sound scenes (S5) involves the accurate identification of active sound classes and the precise separation of their sources from complex acoustic mixtures. Conventional systems rely on a two-stage pipeline - audio tagging followed by label-conditioned source separation - but are often constrained by the absence of fine-grained temporal information critical for effective separation. In this work, we address this limitation by introducing a novel approach for S5 that enhances the synergy between the event detection and source separation stages. Our key contributions are threefold. First, we fine-tune a pre-trained Transformer to detect active sound classes. Second, we utilize a separate instance of this fine-tuned Transformer to perform sound event detection (SED), providing the separation module with detailed, time-varying guidance. Third, we implement an iterative refinement mechanism that progressively enhances separation quality by recursively reusing the separator's output from previous iterations. These advancements lead to significant improvements in both audio tagging and source separation performance, as demonstrated by our system's second-place finish in Task 4 of the DCASE Challenge 2025. Our implementation and model checkpoints are available in our GitHub repository: https://github.com/theMoro/dcase25task4 .