Abstract
An improved DTW algorithm is proposed by studying the features of dynamic gesture track. Using Kinect sensor to obtain the fingertip position in time, construct the angle vector feature to describe the movement track of the fingertip. Relaxation the endpoint limit resolves the problem that the gesture sequence is not aligned with the candidate sequence endpoint, an improved ET-DTW algorithm is obtained by using the early termination matching strategy to reduce the unnecessary amount of computation. At the same time, the candidate sequences were sorted by LB_Keogh lower bound function to further optimize the algorithm. Experiments show that the accuracy and recognition speed of dynamic gesture recognition are greatly improved compared with the traditional DTW algorithm.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call