混沌优化的仓储害虫声音信号识别研究

Storedproducted Insects's Voice Signals Classification Based on the Chaos Optimization

为了提高仓储物害虫声音信号的自动识别率,寻找支持向量机模型参数C和核宽度参数σ的最优组合,提出了基于混沌优化的支持向量机模型参数自动选择算法.基于径向基核函数(Radial Basis Function,RBF)的支持向量机模型参数C和核宽度参数σ对其泛化能力有很大的影响,首先产生Logistic映射和圆映射的混沌混沌数值序列,而后以通过载波形式将混沌变量的值域"放大"至参数(C,σ)的取值空间,寻找优化变量(C,σ)的最优组合.与网格法的比较实验结果表明,该方法不但可以提高分类识别率,而且显著减少了支持向量机的训练个数,并使支持向量机具有更好的推广能力.

In order to improve the automatic recognition rate of voice signals with respect to stored product insects,this paper proposed support vector machine model parameters automatic selection algorithm based on chaotic data series to search for the optimal combination of support vector machine model parameters C and RBF kernel width parameters σ.The parameters（C,σ） have a great impact on the generalization ability.Based on the chaotic sequence generated by Logistic Map and Circle Map,respectively,the chaotic variables are zoomed to the parameter（C,σ） value space to search for the optimal combination.Compared with the grid method,experimental results show that this method can improve the classification rate,and significantly reduce the number of training support vector machines,so the support vector machine has better generalization ability.