中文摘要：

在多目标识别中,决策导向无环图支持向量机（DDAGSVM）是一种有效的方法。但在分类过程中它存在误差累积现象。在分析此问题的基础上,借用广义核函数fisher最佳鉴别的思想,提出了一种基于Fisher判别率的改进DDAGSVM。最后应用改进算法对四种防空战场目标的声信号进行分类识别。实验结果表明它很大程度上降低了累积误差,较原算法提高了分类精度。

英文摘要：

To solve multi-class classification problem, Decision Directed Acyclic Graph algorithm support vector machine （DDAGSVM） has been proposed. But it did not solve the problem of minimizing the classification error that might be accumulated at the final classification process. In order to minimize the classification error, the efficient method is studied in this paper. A separability meas- ure is defined based on Fisher discriminant criterion. And an improved DDAGSVM is presented by introducing the defined between-class separability measure into the formation of DDAG. In the end, the improved algorithm is applied to four categories air defense battlefield acoustic target recognition. And the results proved the effectiveness of the improved algorithms.