Sound Class
Sound class research focuses on developing robust and adaptable systems for classifying and processing audio, addressing challenges like incremental learning of new sounds and the detection of manipulated audio. Current efforts utilize deep learning models, often employing class-incremental learning techniques and metric learning approaches, to improve accuracy and efficiency while minimizing "catastrophic forgetting" when adding new sound categories. This work is crucial for applications ranging from environmental monitoring and assistive technologies to security systems and multimedia content creation, improving the ability of machines to understand and interact with the acoustic world.
Papers
September 11, 2024
July 4, 2024
April 26, 2024
January 17, 2024
November 1, 2023
July 21, 2022
July 15, 2022
May 15, 2022