Peter Flom wrote: > Thanks to those who already replied! > > Their replies make me realize I needed to post a few more details. > > Dear Peter,
in addition to the reply from Carlos A. Estombelo Montesco ... you are working in the area where a lot ot efforts were done - so you will use specific methods desribed in the literature - since I think they are more adequate. But from the common poinf of view, the main problem is to separate some common properties from the specific to your needs (select features) before the classification, otherwise you can classify some trivial subject properties ... i'm sure (since the signals are rather quasi-periodic than exactly periodic) that some wavelet analysis will help you to select features, and i'm sure there are some specific wavelets to be used in this area (like some Gabor packets or similar) - you need to search libraries. Hope this will help you :-)) Anatoly Saveliev, Kazan State Univ., Russia > Here's the problem: > > I will be dealing with EEG data (brain waves). There will be, for each > subject, 19 time series, each with thousands of points (sampling is once > per milisecond). > > We have some old data, which are already classified, but those classes > aren't absolutely certain to be correct. There may be some errors, and > there may be classes that have not been identified in the literature. > > We will be getting new data, where the classes will be unknown. > > We want to look at new data and classify them; we also want to look at > all the data (old and new) for anything previous researchers have > missed. > > > Any further suggestions will be very welcome > > > ---------------------------------------------- CLASS-L list. Instructions: http://www.classification-society.org/csna/lists.html#class-l ---------------------------------------------- CLASS-L list. Instructions: http://www.classification-society.org/csna/lists.html#class-l
