Are all BCI feature extraction methods subject_independent?

A Brain Computer Interface (BCI) utilizes signals derived from electroencephalography (EEG) to establish a connection between a person´s state of mind and a computer-based signal processing system which interprets the EEG signals. Extracting appropriate features from available EEG signals is essential for good BCI communication and an acceptable level of accuracy. Till now, many different feature extraction techniques have been used. Recently, a new set of features called Complex Band Power (CBP) are introduced. In this study(Townsend et al, 2006), showed that CBP features could result in more accuracy in comparison to traditional band power features and Common Spatial Patterns (CSP) features. In this paper, the resulted accuracy from CBP, CSP and traditional band power features were compared using the data set Bci-Competition2005. The simulation results showed the superiority of CBP features over traditional band power features and also showed that CSP features lead to more accuracy (inverse result in compare of previous work). The results indicated that the success of these feature extraction methods depends strongly on the subject and personal differences such as mental patterns and IQ. Both CSP and CBP are powerful feature extraction methods and it is hard to choose one as more appropriate.