Cognitive capability identification in performing mental tasks using EEG-based coherence

Previous research in the field of cognitive science clearly emphasizes the importance of coherence in language processing and the analysis of mental tasks. In this paper, electroencephalography (EEG)-based coherence between different pairs of electrodes has been used for the classification of mental arithmetic capability for different subjects. EEG signals were obtained using 19 electrodes when 36 subjects performed mental arithmetic operations. These EEG signals were denoised using wavelet-based techniques. Then the signals were decomposed into alpha, beta, gamma, delta, and theta frequency bands. The magnitude squared coherence in all the individual frequency bands for different pairs of electrodes was calculated. The high coherence was prevalent in the anterior frontal and frontal electrodes. It can also be seen from this work that the alpha band provides maximum coherence. The coherence features were classified in the alpha band using a non-linear support vector machine and 97.6% accuracy was achieved. To reinforce our findings, the work has been compared in a concurrent framework using statistical features such as mean, variance, and skewness. The classification accuracies were 72.3%, 61.4%, and 53.9% respectively using the above three features respectively. This study shows the effectiveness of coherence features by providing additional insights regarding the involvement of different brain areas in cognitive processes.