I am using 5 channels [ fz , cz , c3 , c4 , pz] to detect drowsiness of driver My First Question is, what is the right input to get feature power band ( Theta , alpha , gamma , beta ) to wavelet transform ? ( these 5 channels or 1 channel or what ? ) My Second Question is, Is it right to classify data based on theta only got from wavelet transform ?
Generally, alpha-band oscillatory activations (8-10Hz) relate to relaxation and in principle accompanied with closure of eyes. This is the prime marker that is used to detect drowsiness, but surely not the only one (see alpha dropout, NREM1, eye-rolling EEG artefacts).
To detect alpha, electrodes from occipital or adjacent regions are used. Due to low SNR in EEG signals and its variance across EEG systems, you may want to try different detection methods to decide for your classifier.