基于参数优化的支持向量机战场多目标声识别

Multi-target Acoustic Identification in Battlefield Based on SVM of Parameter Optimization

针对网格搜索支持向量机(SVM)参数的方法存在复杂度高、运算量大等不足,提出了一种改进的网格搜索SVM分类器的最佳参数选择算法。将其应用于战场多目标SVM分类器中,对该分类器与KNN分类器和改进BP神经网络分类器进行的分类对比实验表明,改进的网格搜索SVM分类器参数选择算法可以有效地减少SVM分类器的运算量、改进学习性能并提高识别率。

The original grid-search algorithm for choosing parameters of Support Vector Machine （SVM） has large amount of calculation in the training processes. An improved grid-search algorithm is proposed to choose the optimal parameters of SVM. The battlefield multi-target SVM classifier is designed using this algorithm. Target classification experiments are done using K-nearest neighborhood classifier, improved BP neural network classifier and SVM classifier respectively. The result shows that the improved grid-search algorithm can reduce the SVM classifier＇s computation effectively and improve its performance and classification accuracy.