Frequency-amplitude coupling: a new approach for decoding of attended features in covert visual attention task

A method that has recently been mentioned as information encoding brain is cross-frequency coupling (CFC). It is generally assumed that CFC can play a crucial role in perception, memory, and attention. In this study, two new indices for evaluating frequency-amplitude coupling (FAC) through generalized linear model (GLM) and linear regression method were introduced and investigated along with other CFC indices. Electroencephalogram (EEG) signals were recorded during covert visual attention tasks to find out the CFC index capability so as to distinguish different states in the mentioned tasks. To this end, machine learning algorithms were used and four various types of CFC, phase-amplitude coupling (PAC), phase-phase coupling (PPC), amplitude-amplitude coupling (AAC), and frequency-amplitude coupling (FAC) in recorded signals were considered as inputs for classifiers. The results demonstrated that the proposed method used for evaluating FAC through linear regression can provide more information about the different states in two covert attention tasks using quadratic discriminant analysis (QDA) by classification performance of 94.21% and 90.54% in color and direction tasks, respectively. Also, FAC that used a GLM model and PAC had a higher performance compared with PPC and AAC in color task (90.74 and 92.24% against 83.21 and 86.22). We can conclude that CFC can encompass useful information about semantic category of stimuli in covert attention tasks and can be used as an acceptable alternative for the time-frequency features of brain signals.