EIQ: EEG based IQ test using wavelet packet transform and hierarchical extreme learning machine.

Background: The use of electroencephalography has been perpetually incrementing and has numerous applications such as clinical and psychiatric studies, social interactions, brain computer interface etc. Intelligence has baffled us for centuries, and we have attempted to quantify using EEG signals.
New Method: This paper aims at devising a novel non-invasive method of measuring human intelligence. A newly devised scoring scheme is used to ultimately generate a score for the subjects. Wavelet packet transform approach for feature extraction is applied to 5 channel EEG data. This approach uses db-8 as the mother wavelet. Hierarchical extreme learning machine is used for classification of the EEG signals.
Result: 80.00% training accuracy and 73.33% testing accuracy was measured for the classifier. The average sensitivity and specificity across all three classes was measured to be 0.8133 and 0.8923 respectively. An aggregate score was determined from the classification of EEG data. The power spectral analysis of the EEG data was conducted and regions of the brain responsible for various activities was confirmed. In the memory test, theta and beta bands exhibit high power, for arithmetic test, alpha and beta bands are strong, whereas in linguistic test, theta, alpha and beta bands are equally strong.
Comparison: The traditional IQ test determines intelligence indirectly, based on the score obtained from Wechsler test. In this paper an attempt is made to measure intelligence based on various brain activities - memory, arithmetic, linguistic.
Conclusion: A new method to measure intelligence using direct approach by classifying the EEG signals is proposed.
(Copyright © 2019 Elsevier B.V. All rights reserved.)