基于自适应小波神经网络的复杂系统模式识别方法

针对传统神经网络应用于复杂系统建模和辨识中存在的训练效率、精度瓶颈问题,提出了一种自适应小波神经网络方法(adaptive wavelet neural network,AWNN).首先,通过设计自适应层、综合层,使神经网络能根据待处理的系统的样本数据特征自适应工作于最佳工作区间;然后,通过将小波分析方法与对经典的基于误差反向传播算法的神经网络(back propagation neural network,BPNN)、径向基神经网络(radical basis function neural network,RBFNN)结合,保留了上述方法的优点,克服了传统神经网络方法各自的问题;最后,通过对BPNN、RBFNN和AWNN方法进行计算机仿真实验,验证了各算法的可行性、可达性和算法参数特性.实验结果表明:AWNN方法具有更快的收敛速度、更高的精度和更好的鲁棒性.

Complex System Pattern Recognition Method Based on Adaptive Wavelet Neural Network

To solve the training efficiency and accuracy bottleneck problems of the traditional neural network method in modeling and identification application of complex systems,an adapted wavelet neural network( AWNN) method was proposed. First,adapted and integrated layers were design to make AWNN create normalization parameter to adapt the sample data. AWNN absorbed the advantages of BP neural network,RBF neural network and wavelet analysis algorithm overcome the problems of the original neural network. A large number of experiments and comparative analysis had been implemented to verify the performance and characteristics of the AWNN. Both computer simulation results and intelligent video analysis application experiments show that the AWNN method has faster convergence speed,higher accuracy and better robustness.