Machine Listening

Machine listening focuses on developing computational systems that can analyze and interpret audio data, mirroring human auditory perception and understanding. Current research emphasizes the development of robust, reference-free audio quality assessment models, such as generative machine listeners, which leverage neural networks and transfer learning to predict subjective listening experiences without needing a reference audio signal. A significant challenge lies in creating high-quality, contextually labeled datasets, often addressed through active learning strategies that combine human expertise with AI-driven sample selection. These advancements are improving audio quality monitoring, sound scene recognition, and other applications, while also offering insights into human auditory processing.