基于证据数据分类算法的水声目标识别研究

The Underwater Acoustic Target Recognition Based on Evidence Classification Algorithm

证据分类算法已被广泛应用于模式识别中。针对传统证据近邻算法在证据权重和组合规则上的局限,研究了一种新的基于DSmT的证据K近邻识别算法(DSmT-KNN)。首先在水声目标的各类别训练模板库中,利用目标数据与各近邻的特征相似度来分别构造基本置信指派,并根据K个近邻数据的距离大小对构造的置信指派进行加权。然后利用DSmT规则对加权证据进行融合。最后根据每个类别下融合结果的算术平均值来判断目标的类别属性。通过水声目标实测数据实验,将DSmT-KNN与其他几种常见的方法进行了对比分析,结果表明新算法能有效提高系统的识别准确率。

The evidence classification algorithm has been widely used in the pattern recognition field. In order to effectively overcome the limitations of the traditional evidence nearest neighbor classification algorithm,a new evidence K-nearest neighbor recognition algorithm based on DSmT（ DSmT-KNN） is presented. In this new method,the basic belief assignments（ bba＇s） are determined using the feature similarity between the object and its K nearest neighbor in each class of the training underwater acoustic targets,and then the K bba＇s are discounted according to the feature distance of the K nearest neighbors. Finally the discounted bba＇s are combined using DSmT rule,and the mean of these combination results in each training class is used for the recognition of the object. Several experiments based on real underwater acoustic data sets are given to test the effectiveness of DSmT-KNN with respect to some other methods. The results indicate that DSmT-KNN can effectively improve the recognition accuracy.