What features could be extracted of EEG dataset?

I want to classify two kinds of EEG data and I used some time domain and frequency domain features, but it could not work well. The accuarcy is only 50% (the training sample number is 70 and testing: 10).

Some papers said that DCT features work well on EEG classification, someone please tell which features could extracted from EEG by DCT? Or other features work well in EEG classification?

Thanks very much.