One of the accepted Grand Challenges of science is to understand the workings of the human brain. This is a problem that has many aspects and many levels, but one of the most promising is to study the brain as a computational system, and in particular to study the lower level processing components, such as early vision and sensory-motor integration which are both experimentally accessible and lend themselves to algorithmic description. The development of such biologically validated algorithms also allows their transfer to build functional computer artefacts mimicking natural behaviours and supporting new forms of human computer interactions.
My work on early visual processing over the past 13 years has shown that a single gradient-based mechanism can account for a range of visual phenomena that were previously attributed to separate neural mechanisms. This computational framework has been psychophysically validated and has lead to the discovery of a number of new visual illusions. This gradient approach has been shown to transfer between visual sub-domains, motion and spatial orientation to date, so providing a contended for a generalised framework for understanding human visual processing. The algorithms developed have then been used to develop robust real time computer vision systems to extract important scene properties such as optical flow and slope.
My work on sensory motor integration examined initially a novel insect behaviour, active motion camouflage, and developed the first explicit neural model for this stealth strategy. Using the model developed it was possible to incorporate it into a virtual environment and show for the first time that humans are also susceptible to this camouflaging behaviour. Examination of human sensory motor performance also allowed the development of an award winning Internet based security system using computer mouse based signatures, which we showed for the first time to be both reproducible and biometric.
Closing the perception action loop in human computer interaction is an extension and integration of this previous work, using biologically inspired methods and computer vision techniques to develop systems capable of tracking faces, recognising facial expressions and so supporting affective computing applications. This work is extending currently to include the automated recovery of body language in an effort to improve interaction. Through such systems we can attempt in the near future to understand many of the higher-level cognitive processes in the human brain concerned with affective communications, social intelligence and empathy.
Research on understanding the brain and the development of artificial intelligence has been identified by the UK Government as an issue that the public have concerns about. Key to the future acceptability of such systems, and continued funding from the public purse of research, is the need for public dialogue on the issues surrounding artificial intelligences. Integral to my research agenda is a successful and ongoing engagement with the public and media to demystify the subject and inspire the next generation of researchers. An example being my collaborative Sodarace project, which was selected for the Royal Society Summer Exhibition 2005 and the cs4fn project.
1997 -2000 BBSRC/EPSRC Mathematical Biology Initiative Project Grant A DIFFERENTIAL FORMS APPROACH TO MODELLING OPTIC FLOW
1999-2002 EPSRC Project Grant. EVOLUTION OF SENSORIMOTOR NEURAL NETWORKS TO PERFORM ACTIVE MOTION CAMOUFLAGE