基于遗传算法的人体穴位阻抗特征优化

Feature optimization method in impedance signal of human meridian based on genetic algorithm

为消除人体穴位的阻抗信号特征集中存在的冗余和不相关分量的问题,提出了一种基于遗传算法的人体穴位阻抗特征子集选择与优化算法.通过分析穴位阻抗信号的自回归(AR)模型谱图建立了穴位原始特征样本集,利用类内-类间距离判据构造遗传算法的适应度函数并改进遗传算法的特征优化算子.经人体穴位的电阻抗特征选择与优化实例分析表明:该方法具有较好的寻优性能和适应度稳定,在不增加原始信息的情况下,能够有效地减少分类识别的特征数和提高信号识别的准确率,且将穴位阻抗特征的平均状态辨识率提高9%左右.

In order to solve the redundant and irrelevant components problem in the human meridian impedance feature set,a selection and optimization method of the human meridian feature subset based on genetic algorithm was proposed.Original meridian sample feature set was established through analyzing autoregressive（AR） power spectrum of human meridian impedance signal.By using Euclidean distance among all instances of different class,the fitness function of genetic algorithm（GA） was constructed and feature-optimization operators of GA was improved.Case study on the feature selection and optimization of human acupuncture points shows that the method has good optimization performance and stability in the fitness function,and the average status recognition rate of human meridian impedance characteristic increases nearly 9% without increasing the original signal information.