On 20/02/06, Florian Lohmann <mail@...> wrote:
> therefore I would like to know if anyone has experiences with
> specific values that correspond to a persons state of mind at the
> time of them being connected to the modEEG and reading them out in
> BioEra, like their state of relaxation, their attention, their focus,
> or their anxiety, and transform those into minimal or maximal or
> AVERAGE values (threshold) over a long period of time (e.g.10
> minutes), or any values that tell me something about the actual state
> of mind user. To be honest, i do not need to have an actual
> scientifically proven meaning on any of those values (although any
> hint of what it might correlate to is highly appreciated), I just
> need maybe 10 FIXED NUMERIC VALUES that are fairly different in any
> person at any time, being sort of a momentary snapshot of the person,
> classifying them and being significantly different.
> Like, if somebody is relaxed, a low number, and if somebody is
> anxious or stressed a high number (...)
> does anybody have a good approach?
I am not really sure what exactly you are trying to achieve but I
believe the easiest approach would be to weight the power observed in
different frequency bands.
has a fairly simple explanation of the interpretation of the different
frequency bands commonly observed in EEG.
So I suppose you should sum up the FFT coefficients found in the
different bands you are interested in (possibly weighted by a
windowing function), multiply each by some factor and add them up to
get your numeric value.
However, if you are trying to do a classification why do you want to
reduce your data to one dimension before you use some classifier? It
may be significantly easier to classify your measurements if you can
base your decision on more than one dimension.
I think an even more simplistic approach would simply detect the
highest power in frequency space and then use a proportional mapping:
low frequency <-> low stress level and high frequency <-> high stress
> And Jarek, excuse my direct approach, but how would i have to build a
> bioera-design which generates a NUMERIC average or a threshold min or
> max from those significant values over a long time period (10-30
> minutes?) Is there a way of automatically counting e.g. heartbeats
> per minute?
You could try to use a correlation filter.