A Visual Hand Motion Detection Algorithm for Wheelchair Motion

This paper describes an algorithm for a visual human-machine interface that infers a person’s intention from the motion of the hand. The context for which this solution is intended is that of wheelchair bound individuals whose intentions of interest are the direction and speed variation of the wheelchair indicated by a video sequence of the hand in rotation and in vertical motion respectively. For speed variation recognition, a symmetry based approach is used where the center of gravity of the resulting symmetry curve indicates the progressive position of the hand. For direction recognition, non-linear classification methods are used on the statistics of the symmetry curve. Results show that the symmetry property of the hand in both motions can serve as an intent indicator when a sequence of fifteen consecutive frames is used for recognition. This paper also shows less satisfactory results when fewer frames are used as an attempt to achieve faster recognition, and proposes a Brute force extrapolation algorithm to better the results.