Real-Time Dynamic Gesture Recognition Based On Boundary-Constraint Dynamic Time Warping

Natural human-computer interaction technology based on gestures has received broad attentions in many fields recently, which requires better performance of gesture recognition algorithms. Real-time gesture recognition is challenging due to the idea of dynamic programming. To address this issue, in this paper, we propose a real-time dynamic gesture recognition algorithm based on boundary-constraint dynamic time warping (BC-DTW). The proposed algorithm reduces the recognition time from the calculation times and the amount of DTW distance. To reduce the calculation times of varies DTW distance between the gesture sequences, the proposed algorithm transforms a gesture sequence from two-dimensional of unequal-length to one-dimensional of equal-length. Moreover, to find the paired range of each gesture data in a sequence during the process of DTW distance calculation, we propose a method based on Freeman chain to determine the boundary width of gesture data. The experimental results show that the proposed BC-DTW algorithm can speed up the recognition compared with the dynamic gesture recognition algorithm based on DTW.