一种基于数据手套的静态手势识别方法

A Static Gesture Recognition Method Based on Data Glove

在虚拟环境的交互任务实施过程中,手势识别的正确率和效率将直接影响到操作者的沉浸感和成功率.针对已有的手势识别方法难以在既保证较高识别正确率的同时又满足实时性要求的问题,提出一种能够有效用于静态手势识别的手型特征提取方法以及相应的手型特征点集匹配策略.首先,利用可穿戴式设备——数据手套采集多种原始的手部运动数据,对这些数据进行预处理后构建出手势库;然后,提取并表示每一种手势的手型特征;最后,运用特征点集模板匹配方法进行手势识别.实验结果证明,该方法在手势类别数目较大(25类手势)时识别正确率能够达到98.9%,并且计算量小、效率高,能够保证用户和虚拟环境交互的实时性.

In the process of interacting with the virtual environment, the gesture recognition accuracy and effi-ciency will directly influence the operator’s sense of immersion and success rate. The available methods are hard-er to keep greater accuracy, and meet the real-time request at the same time. Aiming at this problem, a gesture recognition method based on the shape of the hand feature is proposed in this paper. Firstly, we capture a serial of original data using data glove and construct the gesture base by modifying the original gesture. Then we extract the hand-type characteristic of each gesture. At last, we make the gesture recognition by the improved feature point set template match algorithm. Experimental results show that 98.9% accuracy is achieved when the number of categories of gesture is larger (up to twenty-five Categories). Moreover, the proposed method is simple enough to meet the real-time requirements.