Thesis Project Form

Early interactions between infants and parents are a key component of emotional, cognitive, and communicative development. Differences in the way children engage with parents may reflect early signs of atypical developmental trajectories. For this reason, the observation of parent–child interactions is widely used in developmental and clinical research. Despite its importance, the assessment of these interactions is still largely based on qualitative observation by trained clinicians or manual annotation of video recordings. These approaches can be time-consuming and difficult to standardize, limiting their scalability and reproducibility. Recent advances in computer vision and multi-camera sensing technologies enable the objective and quantitative analysis of social interactions. Multi-view video systems allow interactions to be captured from different perspectives and support the reconstruction of spatial relationships between participants, enabling the extraction of behavioural indicators such as body pose, gaze direction, movement patterns, and interpersonal distance.
Within this context, the aim of this thesis is to investigate how video-based multi-camera systems can be used to capture and analyse behavioural dynamics during parent-child interactions in a controlled clinical environment.

The objective of this thesis is to develop a multi-camera video-based framework for studying behavioural interactions between children and adults.
The work will focus on the design of a video acquisition setup and on the development of computational methods to extract quantitative descriptors of interaction dynamics.
The main activities of the thesis include:
• reviewing existing approaches for video-based human behaviour analysis
• designing and implementing a synchronized multi-camera recording setup in a controlled environment
• developing algorithms to analyse behavioural cues such as body pose, gaze direction, relative orientation, proximity, and gestures
• collecting pilot recordings of parent–child interactions

During the thesis the student will gain experience in:
• configuring and calibrating multi-camera video acquisition systems
• implementing computer vision and machine learning methods for behavioural analysis
• developing software pipelines for video data processing
• designing and conducting pilot experiments involving human participants in a clinical environment