Ronald Poppe


Function:Assistant Professor (PNUT)

Research interests

In my research, I combine my technical background in computer science with my fascination for human behavior. The complexity of human (nonverbal) behavior motivates my research in two ways. First, the focus is on the automatic analysis and understanding of human behavior from video and other sensors. For this, I use and develop algorithms from the fields of computer vision and pattern recognition. Second, the aim is to use insights gained from this analysis to generate nonverbal behavior for virtual characters and robots.

The behavior theories are grounded in psychology, while behavior observations are typically analyzed using computer vision and pattern recognition methods. Taking a multi-disciplinary approach can mutually benefit these fields. On the one hand, formulation and verification of hypotheses on human behavior can benefit from automatic measurement of human motion. On the other hand, known patterns in behavior aid in the understanding and subsequent modeling of nonverbal communicative behavior.

As an example, we are developing methodologies to simultaneously learn and evaluate human behavior generation models. We are doing this in the context of speaker-listener interactions, where a human speaker has a conversation with a virtual listener. We try to learn the quantity, timing and type of listener reponses ("uh-huh", nods) from human-human conversations, but also from interactions with the virtual listener. For the latter, we use peception experiments and online machine learning methods. Future work will address less-constrained conversational settings such as structured interviews and question-answering. We aim to develop more human-like virtual characters such as virtual coaches, (tour) guides and receptionists.

Another current topic we are working on is the automatic analysis of behavior of groups of children during play. This research is being carried out in the context of interactive digital playgrounds, open spaces equiped with sensors and actuators. By facilitating interactions between (groups of) children and between children and environment, we intend to provide a more engaging experience, and at the same time steer the interactions to promote exploration and suppress negative aspects of play such as aggression and isolation.