优化预处理方法提高BP网络模式识别能力

通过改进样本的预处理方式，将神经网络理论中的以分量为单元的预处理方式改为以样本为单元的预处理方式，对样本进行归一化或正规化处理。简化了预处理程序，提高了样本类的可分性，保持了样本各特征量之间的特征关系。实验表明，网络的训练效率和模式识别能力得到大幅度提高。

Improvement of BP network s' classifying ability by changing method of preprocessing

The preprocessing work was simplified, the patterns of signal become more recognizable and the relation between each component of one sample was not changed, if the preprocessing of samples was based on signals rather than on the components of the signals' characteristic vectors. At the same time, neural network can be trained more efficiently and the trained network can classify the inputs in a more appropriate way.