I used that formula to compute the relative power percentages but got different numbers from your online graph. Did you do yours before the filtering/averaging or after the filtering/averaging? These values are not in the csv file so we have to compute them from the absolute eeg data. Also we cannot compute them from the raw samples as the raw samples in the csv are only logged at 1 Hz.
Thank you and please feel free to correct me if I am wrong.
Octave support?
Re: Octave support?
If you want to do your own FFTs, change the recording rate to "Constant" in Settings. This will give you 220Hz RAW EEG data for the original 2014 Muse, and 256Hz for all other models of the Muse. Note this will make the file much much larger and you will get repeated values in the other columns, due to each row requiring a value. You can find the calculation frequencies on the website FAQ section. For example, Absolutes are calculated at 10Hz.
All Absolute values in the CSV are calculated by Interaxon's SDK code. I know they drop blink and jaw clench packets and filter out powerline noise, but it's in the closed source code, so I can't give you their exact calculation to go from RAW to Absolute.
All Absolute values in the CSV are calculated by Interaxon's SDK code. I know they drop blink and jaw clench packets and filter out powerline noise, but it's in the closed source code, so I can't give you their exact calculation to go from RAW to Absolute.
Re: Octave support?
Thanks. I found the "recording interval" "Constant" option and just did a 20 min session. The zip file is about 8 MB, and after unzip it is about 126 MB 
Yes only the raw data changes for every row and others are repeated for multiple rows which helps compress the files. Actually I guess we can just sample at 88 Hz to get the same result as Gamma is only up to 44 Hz.
Sorry I am still not getting the average relative percent as you have for the online graphs. I do alpha=(alpha_1+...alpha_4)/4 first for each row, and then 10^alpha to convert to linear scale, and then sum up all rows to get the total alpha energy. I do this for other waves and then calculate the percent. I am not using the filtered alpha for this calculation. Is this what you do?
Yes only the raw data changes for every row and others are repeated for multiple rows which helps compress the files. Actually I guess we can just sample at 88 Hz to get the same result as Gamma is only up to 44 Hz.Sorry I am still not getting the average relative percent as you have for the online graphs. I do alpha=(alpha_1+...alpha_4)/4 first for each row, and then 10^alpha to convert to linear scale, and then sum up all rows to get the total alpha energy. I do this for other waves and then calculate the percent. I am not using the filtered alpha for this calculation. Is this what you do?
Re: Octave support?
To change the scale from ~[-1:+1] to ~[0:100] you need x=y+1*50
Re: Octave support?
Yeah that's what I tried, but somehow the results are not the same. The averaging/filtering in the log domain will be different from the log of the averaging/filtering in the linear domain. Anyway, both make some senses, though.