We use multimodal behavioural analysis to automatically recognise infant–caregiver attachment patterns from video. The approach models body movement, interpersonal distance and facial behaviour, capturing behavioural cues related to proximity seeking, avoidance and interaction dynamics. Our analysis identifies movement- and distance-based features as particularly informative for attachment classification, while facial Action Units contribute strongly to modelling specific behaviours such as avoidance. The work demonstrates the potential for objective, automated analysis of attachment behaviour, complementing traditional expert coding of infant–caregiver interactions. 

We investigate multimodal stress-detection framework combining physiological signals with behavioural cues from facial expressions, gestures and body posture. The key finding is that body behaviour provides complementary information to facial and physiological signals, with multimodal fusion outperforming existing stress-detection approaches.

Selected Publication: Lin, Weizhe, Orton, Indigo, Li, Qingbiao, Pavarini, Gabriela and Mahmoud, Marwa (2023) Looking at the body: automatic analysis of body gestures and self-adaptors in psychological distress. IEEE Transactions on Affective Computing.