A joint separation-classification model for sound event detection of weakly labelled data

TL;DR

This paper introduces a joint separation-classification model for sound event detection that learns from weakly labeled data, effectively separating sources and detecting events without precise annotations.

Contribution

The paper presents a novel joint model that combines source separation and event classification trained solely on weak labels, improving detection accuracy and source separation quality.

Findings

Achieved an EER of 0.14 in sound event detection, outperforming baseline.

Obtained an SDR of 8.08 dB in source separation, surpassing previous methods.

Demonstrated the effectiveness of GWRP over GMP in probability mapping.

Abstract

Source separation (SS) aims to separate individual sources from an audio recording. Sound event detection (SED) aims to detect sound events from an audio recording. We propose a joint separation-classification (JSC) model trained only on weakly labelled audio data, that is, only the tags of an audio recording are known but the time of the events are unknown. First, we propose a separation mapping from the time-frequency (T-F) representation of an audio to the T-F segmentation masks of the audio events. Second, a classification mapping is built from each T-F segmentation mask to the presence probability of each audio event. In the source separation stage, sources of audio events and time of sound events can be obtained from the T-F segmentation masks. The proposed method achieves an equal error rate (EER) of 0.14 in SED, outperforming deep neural network baseline of 0.29. Source separation SDR of 8.08 dB is obtained by using global weighted rank pooling (GWRP) as probability mapping, outperforming the global max pooling (GMP) based probability mapping giving SDR at 0.03 dB. Source code of our work is published.