Sounds carry a large amount of information about the ambient environment, which allow us to understand and recognize, include the acoustic scenes we are in (office, airport, metro station, etc.), and the events from individual sound sources (dog barking, vacuum cleaner, etc.). Signal processing methods that have the capability to automatically extract this information, have a wide range of application prospects like context-aware devices, multimedia information retrieval, and intelligent acoustic monitoring.

Detection and Classification of Acoustic Scenes and Events (DCASE) is an official IEEE Audio and Acoustic Signal Processing (AASP) challenge, the ninth edition was hold in this year, attracted 123 teams from the industry and academia to participate, received 428 submission entries for all tasks. Built up on the success of the previous editions, this year's evaluation continued supporting the development of computational scene and event analysis methods by comparing different approaches using a common public dataset, continuous effort in this direction to improve the state-of-the-art performance.

The task 4 of DCASE focuses on detection of sound events using real data, with different types of annotations data and corresponding labels available for training. This year it includes two subtasks: Subtask A - Sound Event Detection with Weak Labels and Synthetic Soundscapes, the continuous task of last year; Subtask B - Sound Event Detection with Soft Labels, which is new task to investigate whether using soft labels brings any improvement in performance. Samsung R&D Institute China - Nanjing (SRC-N) won third place at the Subtask A(with ensemble), and second place at Subtask B, which means that Samsung achieved another breakthrough in AI-based acoustic signal processing.

The sound event detection task in DCASE perfectly fits the strategy of SRC-N Intelligence Software Team, which focuses on the sound event in domestic environments. Currently, Samsung TV provides pet care service based on a real-time sound event detection model which developed by SRC-N, and will be promoted to more scenarios in home care domain. SRC-Nanjing Intelligence Software Team focuses on On-Device AI for the goal of providing advanced AI solutions to Samsung device and service. Now the team is engaged in the domains of image quality enhancement, visual understanding, multimodal environment understanding and light-weight AI model development & deployment.

https://dcase.community/challenge2023/indexM
https://dcase.community/challenge2023/task-sound-event-detection-with-weak-labels-and-synthetic-soundscapes-results
https://dcase.community/challenge2023/task-sound-event-detection-with-soft-labels-results