feature exaction

In order to get higher correct rate,three neural networks classifiers with three different feature exaction methods are used separately,then the results are fused by D-S evidence theory to get final decision.
使用了3种不同类型的特征提取方法,分别生成基于神经网络的分类器,并将各自的分类结果通过D-S证据理论进行融合以完成最终的决策,提高了系统的识别率。