EEG analysis of mathematical cognitive function and startle response using single channel electrode

• Published in: CSI TRANSACTIONS ON ICT Vol. 8; no. 4; pp. 367 - 376
• Main Authors: Gopan K, Gopika, Reddy, S. V. R. Aditya, Krishnan, Kumaresh, Rao, Madhav, Sinha, Neelam
• Format: Journal Article
• Language: English
• Published: New Delhi Springer India 01.12.2020
• Subjects: Computer Communication Networks; Computer Science; Input/Output and Data Communications; Original Research; Software Engineering/Programming and Operating Systems; Systems and Data Security; Brodmann Area 10; Statistical analysis; Classificatio; EEG
• ISSN: 2277-9078; 2277-9086
• DOI: 10.1007/s40012-020-00312-5

Electroencephalographic (EEG) signals are non-invasive means of measuring brain functions. EEG has been used in areas ranging from analysis of neurological disorders, emotional states and sleep pattern to Brain Machine Interface. Here we study cognitive functions from a limited set of daily activities of normal human subjects. For this purpose, we utilize EEG signal obtained from prefrontal cortex, specifically Brodmann Area 10L, which is known to play an important role in these functions. In this work, we study characteristics of a task requiring focussed attention (Mathematical Cognition (T3)), an involuntary response to external stimuli (Startle Response (T2)) and the state of rest (Relax (T1)). The single channel EEG is preprocessed and statistical features are extracted from (i) preprocessed data, (ii) wavelet decomposition, and (iii) cepstral analysis. These features are then separately input to various classifiers. Results are reported for 2 class (T1 vs T2, T1 vs T3, T2 vs T3) and 3 class (T1 vs T2 vs T3) classification framework. Experimental results reveal that cepstral analysis is most effective for classification across both frameworks. In 3-class classification, cepstral analysis results in a mean accuracy of 96.61% across classifiers. This framework is shown to be effective in classifying the chosen set of cognitive states using EEG and can be extended to broader classes for more conclusive inferences. In addition, gender differences peak at 81.7%, 77.09% and 77.04% for T1, T2 and T3 respectively, indicating that there are significant differences between the genders, although performing the same cognitive task.