基于ED-DTW算法的手势动作识别方法  

Gesture Action Recognition Method Based on ED-DTW Algorithm

作　　者：张庆宾 吴海波[1] ZHANG Qingbin;WU Haibo(The 28th Research Institute of China Electronics Technology Group Corporation, Nanjing 210007, China)

机构地区：[1]中国电子科技集团公司第二十八研究所,南京210007

出　　处：《指挥信息系统与技术》2018年第1期50-55,共6页Command Information System and Technology

摘　　要：手势动作识别在人机交互和智能应用中发挥着重要作用。针对基于姿态航向参考系统(AHRS)的手势动作识别,提出了一种结合端点检测(ED)及动态时间规整(DTW)算法的手势动作识别方法。首先,根据任务目标定义了5种常用手势动作;然后,对手势动作序列进行ED,使用DTW算法进行模板匹配和分类,从而实现手势动作识别;结合手势动作特征,分别从参考模板建模、失真度计算和最优路径选取3方面提出了更利于提高手势动作识别率的设计原则;最后,试验结果表明,建立的手势动作模板库鲁棒性强,该方法具有较高的手势动作识别率。The gesture action recognition plays a significant role on human-computer interaction and intelligent application.Aimed at the gesture action recognition based on the attitude and heading reference system(AHRS),a gesture action recognition method combined with the endpoint detection(ED)and the dynamic time warping(DTW)algorithm is proposed.Firstly,five common gesture actions are defined by the task target.Then,the gesture action sequence is processed with ED and is classified based on the template matching with the DTW algorithm.Thus,gesture action recognition is realized.With the characteristics of the gesture action,the design principles are presented to enhance the recognition rate of the gesture action from three aspects,including reference template modeling,distortion degree calculation and best path selection.Finally,experimental results show that the template library of the gesture action has strong robustness.Thus,the method has high recognition rate of the gesture action.

关 键 词：姿态航向参考系统 端点检测 动态时间规整算法 手势动作识别 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]