Robust gesture recognition using Kinect: A comparison between DTW and HMM

Human computer interaction through hand gestures is one of the most intuitive ways of communicating with machines and thus it is no surprise that the field of real time gesture detection has seen significant interest among the scientific community in recent times. In this paper a hand gesture recognition method using the Microsoft Kinect has been proposed, which operates robustly in uncontrolled environments and is insensitive to hand variations and distortions. This demonstrates the use of two different learning techniques, dynamic time warping and hidden Markov model and compare them for real-time implementations. The recognition success rate was over 90%. The relative advantages of both techniques have been discussed with constraints.