91É«ÇéƬ

This project will develop multimodal artificial intelligence that enables wearable systems to understand what a user intends to listen to in complex, dynamic environments. Current hearing and wearable assistive technologies mainly process acoustic signals, but do not explicitly model listening intent. The research will jointly analyse eye gaze, head movement, spatial audio, visual context, and wearable EEG, combining behavioural, environmental and neural cues to infer a user’s listening target under uncertainty.

Key outcomes will include multimodal representations of listening intent, uncertainty-aware reasoning algorithms, efficient architectures for real-time wearable inference, and methods for integrating intent understanding with downstream applications such as speech enhancement. The project will also create benchmark datasets, software and evaluation protocols. A research placement at the University of Cambridge will support translation of the developed algorithms into practical wearable platforms, addressing low-latency inference, computational efficiency, energy constraints and robustness to real-world user movement and environmental variability.