粒子群优化神经网络的体育动作识别

Recognition of sports action based on particle swarm optimizing neural network

为了提高体育动作识别的准确性,提出一种粒子群优化神经网络的体育动作识别模型。首先采用背景差法对体育视频图像处理,获得体育动作轮廓,实现体育动作分割,然后提取体育动作的特征,并对特征进行核主成分分析,最后采用BP神经网络对特征向量进行训练,并通过粒子群优化算法选择BP神经网络参数,建立体育动作识别的分类器。测试结果表明,该模型提高了体育动作的识别率,降低了误识率,可以满足体育动作的在线识别要求。

In order to improve the recognition accuracy of sports action, a sports action recognition model based on particle swarm optimizing neural network is proposed. The background subtraction method is used to process the sports video image to obtain the profile of sports action, segment the sports action, and extract the features of sports action. The kernel component analysis is performed for features. The BP neural network is used to train the feature vector. The particle swarm optimization algorithm is used to select the parameters of BP neural network to establish the recognition classifier of sports action. The test results show that the proposed model can improve the recognition rate of sports action and reduce the false recognition rate of sports action, and meet the online recognition requirement of sports action.