基于EEPS-BSS算法的脑深部诱发电位提取

Extracting deep brain evoked potentials based on EEPS-BSS algorithm

脑深部诱发电位对探究神经刺激器治疗疾病的工作机理起着重要作用.为解决临床治疗中采集电极数少于源信号数而产生的欠定盲源分离问题,提出了一种在欠定情况下不需利用先验知识的改进盲源分离法.针对单路脑深部诱发电位的提取,仅利用经验模态分解对观测信号进行分层处理,以一定的规则重构信号来扩展观测信号数目,进而进行分离.仿真与实测数据表明:该方法能实现从低信噪比的单路信号中有效提取微弱诱发电位,分离后各输出信号间的互相关系数较分离前大幅下降,从而证实了该算法提取诱发电位的有效性.

Deep brain evoked potentials play an important role in exploring the mechanism of nerve stimulator to treat diseases. In clinical treatment, there is a problem defined as underdetermined blind source separation that the number of collect electrodes is fewer than the source signal＇s. To solve this problem, an improved blind source separation without a priori knowledge is proposed in underdetermined case. In connection with extracting deep brain evoked potentials from single channel, the empirical mode decomposition is used for stratifying the observed signals, and the observed signals are expanded according to the signals that reconstructed in certain rules. Simulation and measured data show that the novel method could effectively achieve extraction of weak evoked potentials from signals under low signal-to-noise ratio. Compared to unseparated, the cross-correlation coefficients among the respective separated signals are dropped significantly, accordingly, the validity of the algorithm that extracts evoked potentials is confirmed.