A Multiple Camera Approach to Facial Gesture Recognition for Children with Severe Spastic Quadriplegia

Abstract

This paper presents the theoretical framework for a new approach to computer vision-based facial gesture recognition that accommodates the physiological conditions of children with severe spastic quadriplegic cerebral palsy (CP). Clinical observation suggests that these children can exploit one or more facial gestures (e.g. tongue protrusions) to operate a facial gesture access modality with adequate proficiency. The proposed approach uses independent input video data from multiple cameras observing from different viewpoints, in order to maximize the detection of intentional facial gestures in the presence of spastic head movements common to children with severe CP. Also, this paper outlines a case series methodology for further developing and evaluating the proposed algorithm and briefly discusses preliminary image processing issues.