Hedvig Kjellström

The rapid development within computer capacity, in combination with a strong increase in the data available over the internet, has contributed to an avalanche-like development within artificial intelligence. Autonomous cars and systems for smart recommendations of goods from online shops are now commonplace. In order for computers and robots to utilise their full potential when they become more intelligent, they must become better at communicating with humans. One aspect of this is to develop social robots that can behave and react to a human conversational partner’s behaviour in a manner that people perceive to be human and engaging.

Hedvig Kjellström and her research team are developing computer models of how people produce non-verbal communication in the form of body and head movements, facial expressions and glancing directions. They also observe how we perceive and interpret other peoples’ non-verbal cues. This is done using methods from machine learning, which makes it possible to model very complicated connections.
The methods are applied within medical diagnostics (of e.g. Alzheimer’s), on robots who learn from humans, in automatic sign-language interpretation, as well as within theatre and music. The research is interdisciplinary and involves collaborations with doctors, psychologists, musicians and choreographers.