基于改进遗传算法的数字信号特征集选取

摘要

针对通信信号的特点,提出了一种应用于信号特征筛选的改进遗传算法.该方法首先确定了最能表现信号调制间差别的特征子集即优秀基因库,然后在遗传过程中通过选择、淘汰引起优秀基因库大小的变化,最后通过引进不同大小的库外特征量,保证每代遗传过程中的交叉和变异概率随环境的变化而自适应的变化,最终筛选出一高质量的特征子集,并结合RBF神经网络分类器得到更好的识别效果.通过仿真实验验证了该方法不但具有求解全局问题的鲁棒性、收敛性,而且具有更快的收敛速度和更强的全局收敛性.%According to the characteristics of communication signals, a modified genetic algorithm, which can apply to signal feature selection is proposed.This method determines the excellent the gene pool,which can reflect differences of signal modulation. In the genetic process,the steps of selection and elimination cause size change of excellent gene pool. In the genetic process, with the environment changes, crossover and mutation probability of each generation to self-adaptive changes is ensured in the steps. The high quality feature subsets are determined. The RBF neural network classifier is combined to get better recognition result. The simulation results show that the novel algorithm not only has robustness and convergence in solving global problem, but also has faster convergence speed and more stronger or global convergence.