The weights are trained by Kalman filter real-time adjustment technique in the process of sample training, and then the weights are used for check.

• 在样本训练过程中，将卡尔曼滤波递推算法用于神经网络权重的训练，然后用训练得到的权重进行检验。