Sound Event Detection with Sequentially Labelled Data Based on Connectionist Temporal Classification and Unsupervised Clustering
This addresses the problem of reducing annotation costs for sound event detection in audio processing, though it is incremental as it builds on existing CTC and clustering techniques.
The paper tackles sound event detection by using sequentially labelled data (SLD) without timing information, proposing a two-stage method based on connectionist temporal classification and unsupervised clustering. It achieves performance comparable to state-of-the-art systems trained on strongly labelled data and significantly outperforms those trained on weakly labelled data on 41 sound event classes.
Sound event detection (SED) methods typically rely on either strongly labelled data or weakly labelled data. As an alternative, sequentially labelled data (SLD) was proposed. In SLD, the events and the order of events in audio clips are known, without knowing the occurrence time of events. This paper proposes a connectionist temporal classification (CTC) based SED system that uses SLD instead of strongly labelled data, with a novel unsupervised clustering stage. Experiments on 41 classes of sound events show that the proposed two-stage method trained on SLD achieves performance comparable to the previous state-of-the-art SED system trained on strongly labelled data, and is far better than another state-of-the-art SED system trained on weakly labelled data, which indicates the effectiveness of the proposed two-stage method trained on SLD without any onset/offset time of sound events.