Research topics

We combine deep learning, speech/signal processing and computer vision to develop human-centric AI technologies for domains of direct societal importance. Our group is currently working on the following topics:

• Social behaviour analysis: Multimodal emotion analysis, personality assessment, depression detection, anxiety in public speaking, bias in facial expression perception, diversity & inclusion in screen media.

• Trustworthy speech AI: Assessing privacy risks in AI models developed for speech-based affect and behaviour analysis, developing novel attacks and risk mitigation strategies.

• Activity and interaction modelling: Driver activity monitoring, pedestrian and crowd monitoring for smart surveillance, action forecasting, human-human interaction mimicry for human-robot/agent interactions.