Identification of multiple-tasks-induced-EEG by heuristic BCI with learning type fuzzy-template-matching method

Abstract

In the BCI based on the fixed specific EEG features, induced by performing a cognitive task, is suitable for only a particular user, but often the BCI is not suitable for other users. To deal with this matter, we propose heuristic BCI that automatically extracts feature pattern of the brain wave specific to the certain cognitive task performed by a specific individual. In this study, we combined the heuristic BCI with the learning type fuzzy template matching (L-FTM) method. The search space for EEG patterns is composed of the combination of fuzzy labels in antecedent-clause (if-clause) of fuzzy rules, and such specific EEG features are linked to the specific output value of consequent-clauses (then-clause) of the fuzzy rule. In other words, the antecedent-clause of fuzzy rule corresponds to the template for the specific EEG pattern. The values of consequent-clauses are adjusted and fixed by learning process. Thus, learning process is corresponds to searching suitable rules. In this study, we developed a BCI that implements L-FTM method. Inputted feature vector of EEG pattern is compared to each fuzzy rules (templates) and the compatibility value between the rule pattern and inputted EEG pattern is calculated and the output value of the fuzzy reasoning is determined as the average of the values of consequent-clauses of every rules weighted according to the compatibility degree of each rule In addition, we implemented pruning that delete unsuitable rules with high compatibility degree to both of task and non-task status. L-FTM and pruning of the unsuitable rules contributes to the distinction of the EEG expressed during cognitive tasks and EEG expressed during non-task situation. We confirmed the developed BCI with the methods was able to perform the distinction of EEG feature during 2 different cognitive tasks and EEG during non-task situation, even though the program and measurement apparatus were completely same.