基于动态时间规整的手势加速度信号识别

摘要

A dynamic time warping algorithm of gesture recognition was used to improve the performance of dynamic gesture recognition based on accelerometer. The similarity between test template and reference template is calculated by dynamic time warping method(DTW) ,and the results of recognition are got from these similarities. To confirm our algorithm, a wireless gesture acceleration measurement system is proposed and measurements gesture information on 41 volunteers is collected. The results show that the average gesture rate is above 97%. Compared with the algorithm of hidden Markov model(HMM) ,the DTW algorithm is more precise in recognition accuracy.%为了提高基于加速度传感器的动态手势识别算法的性能,本文采用了动态时间规整(DTW)识别算法.通过该算法计算测试模板和参考模板的相似度,从而得出识别结果.为了验证该方法,建立了一套手势加速度无线采集系统,并采集了41个志愿者的手势信息.实验结果表明,该方法手势平均识别率在97％以上.与HMM识别算法相比,DTW识别算法在识别的准确率上比HMM识别算法更具优势.