摘要
在多生物特征融合领域,利用线性加权的方法来进行多分类器融合已有了成熟的应用,但是如何选择一个最优的权值组合仍然是一个值得研究的问题。本文提出了一种利用进化策略来训练分类器的可信度的权值的方法。相比无反馈的权值选取,根据训练样本的识别正确率最大化原则进行训练,最终得到的权值组合再加以识别,可以得到更好的识别效果。仿真实验结果也验证了所提出的方法的有效性。
In the field of multi-biological feature fusion,the use of linear weighting method for multi-classifier fusion has mature applications,but how to choose an optimal weight combination is still an existing problem.Therefore,this paper proposes a method that uses evolutionary strategies to train the weights of the credibility of the classifier.Compared with the selection of weights without feedback,using the principle of maximizing the accuracy of recognition based on the training samples to carry out the final combination of weights for recognition can achieve better recognition results.Simulation experiment results verify the effectiveness of the proposed method.
作者
陈俊
赵子恺
朱梁俊
CHEN Jun;ZHAO Zikai;ZHU Liangjun(College of Mechanical and Electrical Engineering,China Jiliang University,Hangzhou 310018,China)
出处
《智能计算机与应用》
2020年第5期26-29,共4页
Intelligent Computer and Applications
关键词
多生物特征融合
多分类器融合
进化策略
识别正确率
Multi-biological feature fusion
multi-classifier fusion
evolutionary strategy
recognition accuracy
