Enhancement of Filter Design and EEG Power Ratio Features in IQ Pattern Analysis

Power ratio is an established electroencephalogram (EEG) feature that has been used to study cognitive performance. Essentially, the technique computes the normalized power for each of the brainwave components prior to pattern analysis. The method however, is subject to further improvement as previous pre-processing approach rely on low-order filter designs. As a result, the obtained features are less accurate due to the presence of spectral leakages within the pre- processing element. Hence, this paper propose an improved extraction algorithm based on the use of high-order equiripple filters. Pre-existing intelligence quotient data are acquired from 50 samples and their EEG is recorded from the left pre-frontal cortex. The power ratio features are obtained from the energy spectral density of theta, alpha and beta bands. While results maintain conformity with the Neural Efficiency Hypothesis of human intelligence, comparative study shows that with equiripple filters, the revised power ratio is more suitable for IQ pattern analysis.