Here is the .msv file I am decoding
https_nospam_drive.google.com /file/d/0BwLQ6yuHAH3rNnVTQnhPcnNIUi1qZlg5Q20wbkN1ZHd0bVpz/edit?usp=sharing
I am able to load up this EEG file in Matlab
It appears the sampling rate is 200 samples per second
I assume Vsupply is reference (mastoid?), and Va, Vb, Vc are the time series of voltages recorded on particular EEG leads ?
I do not know what 'latency' refers to.
I assume freq indicates the frequency of the vibration tactile stimulus.
I also am not sure what Ta and Tb refer to.
Are the voltage time series (Vb, Vc) already themselves a difference between a scalp lead and a reference lead or do they have to be rereferenced or normalized in some way?
Needless to say, I tried to derive the power spectral density and do not see any peaks at 50Hz, however I also did not see the typical "1/f" curve in the power spectrum, where there is usually a lot of power for low frequencies and small amount of power for higher frequencies,
Could someone provide in layman's terms some insights on what tools I can use to decode this? Are they freely available? What steps should I take from here?
Novice at power spectrum decoding
Re: Novice at power spectrum decoding
It seems Vc and Vd are dead. I'd ignore them.
Your sampling times are not equally spaced. The average sampling rate is just less than 157 Hz. Not many Matlab functions are set up to handle unequal sample spacing. You could ignore it, or you could interpolate the data at a higher rate then subsample the new interpolated data at more consistent sampling intervals.
That should give you a graph of a PSD. I see a peak at ~ 50 Hz.
There's a huge amount of power at f=0 because of the constant offset. If you were to demean/detrend then that would go away.
I've looked at a lot of EEG data and these data do not look very good. Is there any way to get rid of that 50 Hz noise before you record? If it's noise from the vibrotactile stimulator then you'll need to isolate the subject from the device a little better. Are you in Europe? What hardware did you use to acquire these data?
Your sampling times are not equally spaced. The average sampling rate is just less than 157 Hz. Not many Matlab functions are set up to handle unequal sample spacing. You could ignore it, or you could interpolate the data at a higher rate then subsample the new interpolated data at more consistent sampling intervals.
Code: Select all
fs = ceil(1/mean(time));
pwelch(Va, [], [], [], fs);
There's a huge amount of power at f=0 because of the constant offset. If you were to demean/detrend then that would go away.
I've looked at a lot of EEG data and these data do not look very good. Is there any way to get rid of that 50 Hz noise before you record? If it's noise from the vibrotactile stimulator then you'll need to isolate the subject from the device a little better. Are you in Europe? What hardware did you use to acquire these data?
Re: Novice at power spectrum decoding
Thanks for all the information.boulay wrote: If it's noise from the vibrotactile stimulator then you'll need to isolate the subject from the device a little better. Are you in Europe? What hardware did you use to acquire these data?
I'm not sure what hardware we used and I'm still waiting to hear back from the engineer we were with to record it.
I'm slightly embarassed for having supplied you with an auditory stimulus, the purpose of that file was to be the information we "cancel out" from our vibrotactile stimuli. here is a vibrotactile sample.
https_nospam_ drive.google .com/#folders/0BwLQ6yuHAH3rVlJXd21rVnRvX1k
Who is online
Users browsing this forum: No registered users and 14 guests