A Study On The Effect Of ERP Features Evoked By The Mental Arithmetical Loads

Objective Event related potential(ERP) is a most important cognitive evoke potential in neuroscience researches. It is frequently used as the signal feature for brain–computer interface(BCI) control. Mental load generated by mental arithmetic tasks can induce ERPs. This dissertation presents three types of mental arithmetic task and investigates the characteristic of ERP signals induced by them, and thus provides theoretical bases and technical supports for ERP-based BCI paradigm designing.Methods Three types of paradigms are designed and programmed in Matlab. EEG data are collected from 8 healthy human subjects who performed three types of mental arithmetic tasks. In time domain, waveform features of evoked ERPs are studied. Granger causality analysis is applied to explore the brain function network connection in the frequency domain. Further, sample entropy method is applied in characteristic signal complexity analysis. The effect and characteristics of the signal induced by three mental arithmetic tasks are investigated. Finally, signal features are classified using support vector machine(SVM) method.Results All three experimental paradigms can induce obvious P300 component in the time domain and P300 component has larger amplitude at central and parietal region.(1)The difference in P300 latency led by different mental load tasks indicates the difference on information processing speed. The first two tasks are relatively simpler, in that way subjects gain higher response speed. Mental arithmetic based on the strokes of Chinese Characters task is more complex, thus subjects take longer to process information, which lengthens P300 latency 50±15ms longer.(2)Simple counting mental arithmetic task induces maximum amplitude with the average 16.3±0.9μV at the beginning of the experiment. Random number mental arithmetic task induces second large amplitude with the average 10.9±0.4μV, and mental arithmetic task based on the strokes of Chinese Characters induced the minimum with the average 8.5±0.2μV. That is, amplitude of P300 decreases as the task difficulty increases. As experiment goes on, subjects obviously get adapted to simple counting task and amplitude of P300 decreases to 66.3% of initial amplitude significantly. random number tasks is down to 82.6% of initial amplitude, While in mental arithmetic based on the strokes of Chinese Characters, amplitude of P300 is relatively stable.(3)In frequency domain complex mental task induces higher mental load, which enhances the connection intensity, makes brain network a higher connection. In addition, network connection intensity in high-frequency section is much higher than that in low-frequency section. More complex task induces frontal parietal occipital region closer network connection.(4)The sample entropy of the signal collected from subjects in non-attention condition is higher than that in attention condition. In attention condition, the brain is in synchronized state, so there are less nonlinear connections between neurons, and therefore the sample entropy of the signal decreases. On the contrary, the brain allocates more resources so to process more complicated works when the subject is under the task of mental arithmetic based on the strokes of Chinese Characters. At this point, nonlinear connections between neurons increase, along with high complexity and large sample entropy.(5)Support vector machine(SVM) method is applied to do the classification. It indicates that with the increase of superposition times on ERP, the classification accuracy improves. SVM based on parameters of sample entropy gains higher classification accuracy than that based on amplitude with the same superposition. 90% classification accuracy can be achieved when superposition of wave amplitude is repeated 10 times, while sample entropy can achieve the same accuracy with only 4 times superposition.Conclusion In conclusion, this study provides three types of mental arithmetic task and experimental paradigm to induce significant event related potential signals and it can achieve high classification accuracy. Therefore, the proposed paradigm based on mental arithmetic task can be applied to ERP-based BCI system.