0
$\begingroup$

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 ?

$\endgroup$
  • $\begingroup$ Your question is quite broad. It starts with a general question regarding wavelet transform, however, it is unclear what kind of wavelet transform you are using, what you have tried yourself so far and so on.The other questions address primary neuro-physiological signal processing in the field of "drowsiness". I highly recommend you to read the literature in this field and familiarize yourself with that kind of data. Most importantly you should understand digital signal processing in general to choose the right processing method and parameters. $\endgroup$ – Irreducible May 18 '18 at 5:54
  • $\begingroup$ I used discrete wavelet transform; wavedec function, db1 , with level 6 to get feature power bands, $\endgroup$ – Nader Elsheikh May 18 '18 at 11:18
  • $\begingroup$ Very broad question. It basically sounds like you want us to solve your thesis problem for you. $\endgroup$ – mathreadler May 20 '18 at 19:50
0
$\begingroup$

Alpha rhythm

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).

Event detection

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.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.