Jianfeng Liu¹, Zhigeng Pan¹, Xiangcheng Li²

¹ State Key Lab of CAD&CD, Zhejiang University,
310027 Hangzhou, China
{liujianfeng, zgpan}@cad.zju.edu.cn

² Sport System Simulation Lab, China Institute of Sports Scinence,
100061 Beijing, China
lxc@3s.org.cn
 

Abstract. This paper proposes an accelerometer-based gesture rec-ognition algorithm. As a pre-process procedure, raw data output by ac-celerometer should be quantized, and then use discrete Hidden Markov Model to train and recognize them. Based upon this recognition algorithm, we treat gesture as a method of human-computer interaction and use it in 3D interaction subsystem in VR system named VDOM by following steps: establish Gesture-Semantic Map, train standard ges-tures, finally do recognition. Experimental results show that the system can recognize input gestures quickly with a reliable recognition rate. The users are able to perform most of the typical interaction tasks in virtual environment by this accelerometer-based device.