基于多模态眼动识别的乒乓球运动员动作预测

Prediction of Table Tennis Players Action Based on Multimodal Eye Recognition

眼动的速度最高可达到600°/s,具有比人手更灵敏的反应,因此有更好的实时性和应激性,通过眼动识别可以准确反映人体的动作特征,实现动作的判断和预测。采用眼动识别的图像处理算法实现对乒乓球运动员的动作预测,提高攻球效能,改进乒乓球攻击和防御的目的性。传统方法中对乒乓球运动员的眼动识别算法采用边缘特征融合算法,对运动员的动作变换跟踪性能不好,提出一种基于多模态融合眼动识别的乒乓球运动员动作预测算法。提取边缘特征,进行虹膜定位设计,统计搜索区域灰度直方图分布,建立虹膜颜色与边缘联合特征模板,通过多模态融合眼动时变分析,采用从粗到细的处理方法,在减少匹配对应项的同时进行动作预测相关系数匹配,预测跟踪目标中心,实现动作预测,仿真结果表明,算法对运动员的动作预测准确度高,实现对乒乓球运动员动作的实时跟踪和识别。

Eye movement speed can reach a maximum of 600 degrees/s, is more responsive than a human hand, real-time and stress so had better, with eye recognition can accurately reflect the motion characteristics of the human body, the real?ization of judgment and prediction of movement. Using image processing algorithms to achieve eye recognition of table ten?nis athletes action prediction, improve the efficiency of improved objective to attack the ball, table tennis attack and de?fense. The traditional method of eye movements in the recognition algorithm of table tennis athletes using the edge feature fusion algorithm, motion tracking performance of athletes transformation is not good, put forward a prediction algorithm of multimodal fusion eye recognition based on table tennis athletes in action. The edge feature of iris localization design, statis?tical search regional gray histogram distribution, the establishment of a joint feature of iris color and edge template, through multimodal fusion eye time-varying, using the processing method from coarse to fine, in reducing the matching item at the same time should be action prediction correlation coefficient matching, prediction and tracking target center, the realization of motion prediction simulation results show that the algorithm, the action of athletes high prediction accuracy, real-time tracking and recognition of table tennis athletes in action.