基于加速度传感器的大手势集手势识别算法改进研究  

Gesture Recognition Algorithm Improvement with Large Gesture Dictionary Based on Accelerometer

作　　者：王原[1] 汤勇明[1] 王保平[1] 

机构地区：[1]东南大学电子科学与工程学院,南京210096

出　　处：《传感技术学报》2013年第10期1345-1351,共7页Chinese Journal of Sensors and Actuators

基　　金：国家"863"高技术研究项目(2012AA03A302);高等学校学科创新引智计划项目(B07027)

摘　　要：搭载着加速度传感器的智能移动终端为手势识别提供广泛的应用平台,在已有的基于单枚加速度传感器的手势识别研究中,识别正确率、速度和手势集合难以达到良好的平衡。以一种基于加速度符号序列的识别算法为基础,改进了其特征提取方法,并设计了基于加权树结构模板库的匹配方法,实现了大手势集、高正确率、速度快的手势识别。实验表明,所述系统在21个手势组成的手势集合上,实现了95.2%的用户依赖识别率和94.6%的用户非依赖识别率,识别时间小于10 ms,对手势识别研究有一定的借鉴价值。Equipped with accelerometer,smart mobile terminal is the most possible platform for wide application of gesture recognition for UI purpose. In related researches of gesture recognition with single accelerometer,recognition accuracy,speed,and gesture dictionary can hardly be balanced. Based on an algorithm of acceleration sign sequence, this research completes gesture recognition with large gesture dictionary, high accuracy, and speed, by enhancing original feature abstracting method and designing new template match method corresponding to the weighted Tree Structure template base. Experimentally, on a gesture dictionary of 21, the proposed system can achieve a person-dependent accuracy of 95. 2% and a person-independent accuracy of 94. 6%. The recognition time is less than 10 ms.

关 键 词：手势识别 加速度传感器 特征提取 模板匹配 

分 类 号：TM391.4[电气工程—电机]