脑电数据近似熵与样本熵特征对比研究

发布时间：2018-04-22 16:37

本文选题：近似熵 + 样本熵　；参考：《计算机工程与设计》2014年03期

【摘要】：近似熵和样本熵均是量化时间序列复杂性的重要指标,通过两组数据集讨论近似熵与样本熵哪种更适合作为脑电特征。实验结果表明,样本熵作为特征比近似熵能更恰当反映情绪活动存在的差异,差异电极主要集中在大脑前额;以样本熵为分类特征识别嗜酒成瘾者与正常人的平均准确率为80.25%,高于近似熵为分类特征的74.25%,且样本熵算法的计算时间比近似熵算法几乎节约一半。相对于近似熵来说,样本熵更适合作为脑电特征。
[Abstract]:Both approximate entropy and sample entropy are important indexes for quantifying the complexity of time series. This paper discusses which approximate entropy and sample entropy are more suitable for EEG characteristics through two groups of data sets. The experimental results show that the sample entropy as a feature can reflect the difference of emotional activity more appropriately than the approximate entropy, and the difference electrode is mainly concentrated on the forehead of brain. The average accuracy rate of identifying alcoholism addicts and normal people with sample entropy as classification feature is 80.25, which is higher than that with approximate entropy as classification feature 74.25, and the computing time of sample entropy algorithm is nearly half that of approximate entropy algorithm. Compared with approximate entropy, sample entropy is more suitable for EEG features.
【作者单位】：太原理工大学计算机科学与技术学院;北京工业大学国际WIC研究院;
【基金】：国家自然科学基金项目(61070077、61170136) 山西省自然科学基金项目(2010011020-2、2011011015-4) 北京市博士后工作经费基金项目(Q6002020201201)
【分类号】：TN911.7

【参考文献】