基于改进的PCA-PSO-LSSVM乒乓球下旋球发球轨迹研究

Research on serving track of table tennis backspin based on improved PCA-PSO-LSSVM

为了对下旋球发球的整体性进行研究,本文通过标记点对乒乓球运动员关键部位和乒乓球拍进行标记,利用三维运动捕捉系统对整个下旋球发球过程进行研究。针对传统算法中的分类模型适应性差、分类精度不足和LSSVM相关参数的选取随机性强等相关问题,提出了将3种算法结合在一起对乒乓球发球规范性分类的新方法。通过分析下旋球发球的影响因素,选取9个特征向量,并对其进行效果测试,最终确定8个特征向量;通过PCA进一步提取特征向量,把提取结果输入到LSSVM中,用PSO算法优化;根据对比其他几种优化算法,建立了改进的PCA-PSO-LSSVM乒乓球下旋球发球规范预测模型。以两男两女运动员各150组三维下旋球发球数据为例,对模型进行训练和预测,为乒乓球下旋球发球规范性分类提供了一种更为有效的预测方法。

So as to research the integral of the backspin serve,the article utilizes the three-dimensional motion capture system to research the backspin serve by marking the key parts of the table tennis athletes and the paddle.Pointing at the difficulties of bad flexibility of the classification model,insufficient classification accuracy,and casual option of the significant parameters of LSSVM in primitive algorithms.This paper utilizes a new method for the classification of backspin serve standard that unites the three algorithms.By means of the analysis of the useful factors of the backspin serve,9 eigenvectors are selected,and the effect is tested,and 8 eigenvectors are reserved.Then eigenvectors is one step further separated by PCA,and the extraction result is input into LSSVM.In the meantime,by comparison with other majorization algorithms,the PSO algorithm is accustomed to the model parameters’majorization,and an improved PCA-PSO-LSSVM table tennis backspin is established Serving model.Taking 150 sets of three-dimensional backspin serve data of two male and two female athletes as an example,the model is trained and appraised.afford a more successful prediction method for the classification of table tennis backspin serve standard.