基于样本熵的心算任务诱发事件相关电位特征提取的研究

Study on the features extraction of ERP evoked by the mental arithmetic tasks based on the sample entropy

目的利用样本熵研究心算任务诱发的事件相关电位（ERP）特征提取方法，从而增强脑-机接口（BCI）中脑电（EEG）信号成分特征。方法设计计数、随机数和汉字笔画数求和等3种心算认知任务，分别采集3种任务下8名健康受试者16导脑电数据，利用样本熵算法计算脑电特征信号复杂度，探讨心算任务诱发ERP复杂度的特点和不同。结果非靶刺激诱发的信号的样本熵值高于靶刺激诱发信号样本熵值且差异有统计学意义（P〈0．01）；汉字数字笔画心算任务的样本熵值明显高于其他2种任务，差异有统计学意义（P〈0．05）。受试者在非注意状态的非靶刺激下样本熵值高于注意状态的靶刺激的熵值。结论在汉字数字笔画心算任务中，处理的信息复杂，神经细胞间的非线性连接较多，复杂度高。心算任务能有效地激活相关脑区，样本熵可将靶与非靶刺激响应信号相区别。

Objective To study the feature extraction methods for the event related potential （ERP） evoked by mental arithmetical tasks through the sample entropy, in order to enhance the features of electroencephalograph （EEG） signals for brain computer interface （BCI）. Methods Three types of mental arithmetic tasks including a simple counting, a random number and a stroke of Chinese character counting were proposed and 16 channel EEG signals were recorded from eight healthy subjects. The sample entropy method was then applied in characteristic signal complexity analysis. The characteristic and difference of signal complexity of ERP evoked by three types of mental arithmetical tasks were explored. Results The entropy value for EEG signal evoked by non-target stimulus was higher than that by the target stimulus with the significant difference （P〈0.01）. The entropy of the mental arithmetic based on the Chinese characters counting task was significantly higher than that of the other two tasks （P〈0.05）. EEG signals evoked by target/non-target were fundamentally signals under the state of attention or non-attention. Conclusions For the Chinese characters counting task, more complex information have been processed by the brain and the non-linear connection between nerve cells are much more complicated and a higher entropy value is achieved. In summary, the mental arithmetic task can effectively activate the relevant brain regions and the sample entropy can distinguish signals evoked by target or non-target stimuli.