Acoustic(sound) event detection

• The objective of AED is to identify and classify different types of sounds present within an audio signal.

Unsupervised Anomalous Sound Detection

• Our proposed robust training framework for unsupervised anomalous sound detection won the first place in DCASE2022 Task 2!

Sound Event Localization and Detection (SELD)

• In the ICASSP 2022 L3DAS22 Challenge, we utilized a multi-feature fusion deep learning model along with Ambisonics-format audio data augmentation techniques to overcome model overfitting and significantly enhance system performance. The final submitted model achieved a T2Metric score of 0.592, securing second place globally.

  

Audio-Visual Fusion Localization

• Our proposed audio-visual fusion detection and localization framework secured second place globally in the DCASE 2023 Task 3.A Audio-Visual Fusion Sound Event Detection and Localization.

  

• Our proposed audio-visual fusion detection and localization framework achieved fourth place globally in both DCASE 2024 Task 3.A and Task 3.B Audio-Visual Fusion Sound Event Detection and Localization.

  

Low Complexity SED

• Our proposed low-complexity acoustic scene classification training framework achieved third place in DCASE 2025 Task 1！

  

City Scene Perception

• Our proposed city scene perception model achieved third place in the APSIPA ASC 2025 Grand Challenge.！

  

Video demo (Windblade Fault Detection)