# What computerized EEG analysis algorithms are there?

I got a commercial-grade EEG headband and am trying to look at the data it outputs. The headband quantifies raw EEG signal from a single forehead dry sensor into a range of EEG bands (alpha, beta, etc). The issue that I'm running into is that the values of the EEG bands jump rather significantly from one data point to the next (0.05 second reading intervals). Additionally, there are eye-related artifacts.

I would like to apply some sort of algorithm or statistical analysis to see how the EEG bands behave over some time interval. I'm hoping to see if some EEG band statistically rises or is suppressed over 5-15min time period. Seeing instantaneous jumps in individual readings does not convey this information.

For example, I know that for Actigraphic studies, there are sleep scoring algorithms, like one by Cole that looks at the previous 4 minutes, the current minute and 2 minutes in the future to score a minute of actigraphy data:

 SleepScoreConstant*
(550*4minAgo + 378*3minAgo + 413*2minAgo + 699*1minAgo
+1736*CurrentMinute  +


Are there some similar algorithms for EEG band analysis?

Alternatively, it would help if you can suggest the realistic time window over which EEG bands may be statistically analyzed in awake subjects: is it 20 seconds? 5 minutes? 90 minutes?

-
I had commented on one of your questions a while ago that even though the band may be "commercial-grade" you're still basically getting forehead EMG instead of EEG readings. I'll see if I can formulate an answer a bit later on some of the other points, though! –  Chuck Sherrington Feb 17 '13 at 3:20
I came to the same conclusion by looking at the raw output yesterday. Eye movements, motion and jaw motions dwarf the smaller "EEG" signal in a way that reminds me of frequency modulated radio –  Alex Stone Feb 17 '13 at 16:31
Well, even good electrodes pick those things up to some extent, but the better the connection you can get with the scalp (i.e., conductive paste, etc.) the better your chances of getting actual EEG readings. –  Chuck Sherrington Feb 17 '13 at 23:47
(and as an informal answer to your actual question, there are a lot of algorithms to remove eye movement noise from recordings, but you usually need an electrode by the eye to pick up the movements) –  Chuck Sherrington Feb 18 '13 at 1:22
I saw some abstracts of articles on removing motion artifacts, but the actual articles were behind pay walls. –  Alex Stone Feb 20 '13 at 4:00