The Current Feasibility of Gesture Recognition for a Smartphone

The need to improve communication between humans and computers has been instrumental in defining new communi-cation models, and accordingly new ways of interacting with machines. Humans use gestures as a means of communica-tion. The best example of communication through gestures is given by sign languages. With the latest generation of smartphones boasting powerful processors and built-in video cameras it becomes possible to develop complex and compu-tational expensive applications such as a gesture recognition system. We present a gesture recognition prototype system for smart-phones. In order to distinguish between foreground and background objects we describe two segmentation techniques, based on RGB and HSI color spaces. We conclude that the RGB based rules outperform those of the HSI technique. A model based representation alongside a template based technique were developed for hand posture classification. Comparison between methods evidences the latter superior recognition rates. The sequences of hand postures forming gestures were represented through hidden Markov models. We trained a set of three gestures and based on user interac-tion obtained an average recognition rate of 83.3%. We con-cluded that the latest smartphone generation is capable of executing complex image processing applications, with the most penalizing factor being camera performance regarding capture rates