Sound event detection in synthetic domestic environments

Romain Serizel (1) , Nicolas Turpault (1) , Ankit Shah (2) , Justin Salamon (3)

1 MULTISPEECH - Speech Modeling for Facilitating Oral-Based Communication
2 LTI - Language Technologies Institute [Pittsburgh]
3 Adobe Research

We present a comparative analysis of the performance of state-of-the-art sound event detection systems. In particular, we study the robustness of the systems to noise and signal degradation, which is known to impact model generalization. Our analysis is based on the results of task 4 of the DCASE 2019 challenge, where submitted systems were evaluated on, in addition to real-world recordings, a series of synthetic soundscapes that allow us to carefully control for different soundscape characteristics. Our results show that while overall systems exhibit significant improvements compared to previous work, they still suffer from biases that could prevent them from generalizing to real-world scenarios.

Keywords: Sound event detection; synthetic data; weakly labeled data; semi-supervised learning

Domains: Artificial Intelligence [cs.AI]; Machine Learning [cs.LG]; Signal and Image Processing

Romain Serizel, Nicolas Turpault, Ankit Shah, Justin Salamon. Sound event detection in synthetic domestic environments. 2019. ⟨hal-02355573v1⟩