Algorithmic Impromovement Of Dynamic Hand Gesture Recognition Based On HMM Algorithm

Abstract

Compared with other methods of dynamic gesture recognition, dynamic hand gesture recognition based on Hidden Markov Model (HMM) is more widely used. This method aims at analyzing hand signals. In traditional HMM method, any gestures would be calculated with the largest probability as the final recognition result according to Forward - Backward algorithm [1]. In this paper, We study and propose improved algorithm by using some methods like graphics normalizing, probability range limit, number of points limit, coding species limit, etc. to enhance the HMM method for improving the recognition rate and reducing the false detection rate. Experimental results show that the gesture trajectory recognition rate of such method is about 94%, which is higher than the traditional rate of 80%. Furthermore, the excluding error recognition rate, which was not dealt in the past, has reached 100%. Besides, our algorithm can be widely used because of its simple method of equipment, high precision and better robustness.