基于BP神经网络的运动文胸肩带属性与乳房振幅的函数关系

为确定运动文胸肩带的3种属性在人体跑步时对胸部振幅的影响,选取8名被测人员,在其左胸上标记6个测量点,更换不同的肩带进行人体运动测试,记录这些测量点动态的三维坐标,进而得到乳房运动的振幅;利用BP神经网络模型,通过更换不同的网络模型参数,确定运动文胸肩带的3种属性与乳房振幅之间的权值关系。结果表明,选取BP神经网络的传输函数为tansig函数,隐含层神经元个数为21个,训练函数为traingdm作为网络参数时,网络拟合出的乳房振幅值达到了真实值的99.44%;在该网络参数下,分别求得网络输入层到隐含层和隐含层到输出层的权值和阈值,最终得到肩带的3种属性与胸部振幅的正向推理关系式。

Prediction of relationship between shoulder strap attribute and breast amplitude of sports bra by BP neural network

In order to determine the influence of 3 properties of the sports bras shoulder straps on breast amplitudes during running, the motions of breast markers were attached in 6 different left breast positions of 8 subjects, and different shoulder straps were replaced for the human motion testing. The 3-D coordinates of the breast markers were recorded, and the amplitudes of the breast motion were obtained. The weight relationship between the 3 properties of the shoulder straps and the breast amplitude was determined by using the BP neural network model to replace different model parameters. The results show that when the transmission function of BP neural network is tansig function, the number of implicit layer neurons is 21, and the training function is traingdm as the network parameters, the breast amplitude values of the network fitting is close to the true values of the breast motion to 99.44%. The weights and thresholds of the network input to the hidden layer and the hidden to the output layer are respectively obtained under the network parameters. The positive inference relationship between 3 kinds of properties of the shoulder straps and the breast amplitudes can be obtained.