The BCS staged the 2011 Roger Needham Lecture at the Royal Society in London on Tuesday the 1st of November. In this annual lecture the winner of the BCS Roger Needham Award is officially awarded the prize and presents their work to the public. This year’s winner is Prof Maja Pantic from Imperial College London. The lecture on ‘Machine Understanding of Human Behaviour’ presented a decade’s research in the area.
Following an introduction by Prof Jim Norton (president of the BCS), the BCS/CPHC Distinguished Dissertations awardees were announced by Prof Ann Blandford (UCL). The prize was awarded to Daniel Greenfield (Cambridge University); runner up is Vera Demberg-Winterfall (Edinburgh).
The presentation of the Roger Needham Award 2011 to Prof Maja Pantic was carried out by Dr Andrew Blake (Head of Microsoft Research), who identified her a s a driving force in the area of human behaviour recognition. In her excellent lecture on Facial Behaviour Understanding, Maja engrossed the audience with her enthusiasm for the research she has committed herself to. She introduced the audience consisting of academics and representatives from industry to the history of her research in the area of facial expression recognition. This started in an MSc project on static analysis of human facial expressions at Delft. Here, she explored prototypic facial expressions using rule-based systems and was able to distinguish six basic expressions. A total of 45 facial action units directly linked to the contraction of muscles were subsequently established, but not all of these were recognisable with the methods at hand in 2001. The problems related to motion and dynamic expression were not recognisable at the time. This led to the development of a technique called facial point tracking, that has been the focus of Maja’s research over the past 10 years. One of the problems calling for a solution was sudden and drastic head movement. Also, drastic changes in illumination would previously prevent facial expressions from being recognised. Temporal models have been developed to help identify errors due to artifacts by enabling to track actual possibilities given the requirements of the physical muscular movement. Temporal evolution in face videos offers possibilities to detect spontaneous laughter as opposed to acted laughter (real joy versus acted happiness). Affective dimensions of dynamic continuous behaviour then led to multi-dimensional continuous interpretation-space mappings rather than the discretisation used in previous approaches. A new regression method has been established to deal with these.
Maja Pantic concluded her lecture with thanks to her group and all of her previous collaborators, without whom the development of many techniques would not have been possible.
In the question and answer session, Maja competently and enthusiastically answered questions on how successfully people can mask their emotions, the impact of her work in other areas, the extent to which expressions are learnt, the processing power needed for the analysis, and possible extensions of existing speech recognition with her expression recognition techniques? She emphasised that she ultimately wants to help people understand themselves better, in particular persons struggling in social interactions.
The event closed with a vote of thanks by Tom McEwan (Napier University) and a buffet.
The lecture was filmed and is going to be made available on the BCS web site.
(this article will be published in the AISB Quarterly)