一种基于MEMS惯性传感器的手势识别方法

摘要

随着手机等移动电子设备的发展,应用于嵌入式平台的基于MEMS惯性传感器的手势识别成为一个研究热点.提出了一种简单有效的手势识别方法:通过分析手势的运动学特征,在线实时提取手势的加速度和角速度信号特征量,截取手势信号段,利用决策树分类器进行预分类,根据手势信号的变化规律实时识别具体的手势.该方法在20位实验者中获得了96％的平均准确率,手势识别时间小于0.01s.实验结果表明该算法在嵌入式平台下能快速准确地识别手势,满足了实时人机交互的要求.%With the development of smart mobile devices,such as mobile phones,gesture recognition based on MEMS inertial sensor and embedded system has become a research hotspot.A simple but effective gesture recognition method is proposed here.In gesture defined phase,14 gestures are defined,and divided into three categories based on the similarity linguistic and operating.In gesture segmentation phase,the gesture was captured with the kinematic features share by all gestures.In gesture recognized phase,the captured gestures are pre-classified by a two-stage decision tree classifier with the acceleration and angular velocity kinematic features of the various categories of gestures respectively.Experiment among 20 experimenters achieved an average accuracy of 96％.Gesture recognition time is less than 0.01 s.The results show that the proposed method can recognize gestures rapidly and accurately under embedded system and it meet the requirements of real-time human-computer interaction.