基于Jensen熵的运动想象脑电信号稳态子空间分析

Finding stationary subspaces in motor imagery EEG signal based on Jensen-Shannon divergence

目的虽然稳态子空间分析(stationary subspace analysis,SSA)算法在脑电研究领域取得了一定的成效,但目前该算法还不够完善,脑电数据分类误差还比较大,因此要想更好地研究脑电信号,就必须进一步加强算法优化,减少分类误差。本文提出了一种基于Jensen熵(Jensen-Shannon divergence,JSD)的稳态子空间分析算法,并将改进后的算法应用到二类和四类运动想象脑电信号中。方法将JSD代替原SSA算法中的KL散度(Kullback-Leibler divergence,KLD),对改进后的算法(以下简称为JSSA算法)进行模拟仿真,然后将SSA算法和JSSA算法应用到二类和四类运动想象脑电信号中,对Graz2003和Graz2008数据集进行分类提取,并用t检验方法考量SSA算法和JSSA算法所得到的分类准确率是否有显著提高。结果相比于普通算法,SSA算法可以提高运动想象脑电数据的分类准确率,而且基于JSSA算法比基于SSA算法能使运动想象脑电信号分类效果更加准确。结论基于Jensen熵的运动想象脑电信号稳态子空间分析算法相比于SSA算法准确率更好,从而可以使运动想象脑电分类准确率更高。

Objective Although the stationary subspace analysis（ SSA） algorithm in EEG study makes certain achievements,the algorithm is still not perfect. Classification error of EEG data is relatively large,so it is necessary to further strengthen and optimize the algorithm and reduce the classification error. An SSA algorithm based on Jensen Shannon divergence（ JSD） is proposed in this paper,and the improved algorithm is applied to the motor imagery EEG signal. Methods First,we elaborate the principle of SSA and JSD,and then use JSD to replace Kullback-Leibler divergence（ KLD）,and simulate the improved algorithm（ as JSSA algorithm）. The SSA algorithm and JSSA algorithm are applied to motor imagery EEG（ data Graz2003 and Graz2008）,and t test methods are applied for consideration of SSA and JSSAalgorithm＇s classification result which is significant to improve the accuracy. Results Compared with the common algorithm,SSA algorithm can improve the classification accuracy of the motor imagery EEG data,and the algorithm based on JSSA algorithm can make the classification results of movement imagination EEG signal more accurate than based on SSA. Conclusions Compared with the SSA algorithm,the accuracy of the JSD algorithm is better,so that the classification accuracy of motor imagery EEG is higher.