This paper proposed a hybrid feature selector based on ant colony optimization and firefly algorithm. The proposed feature selector is very efficient for selecting features components instead of conventional, and others feature selector. We have modified the support vector machine classification algorithm. The modified support vector machine is called derived support vector machine, and it is multi-level kernel function for the processing of margin. For the process of feature, extraction applied discrete wavelet transform methods. The discrete wavelet transform is derived from the mother wavelet transform. The decomposition of wavelet transforms in due the impulsive noise, and this noise minimized with hybrid feature selector. The proposed system enhances the classification ratio of 8 % with a different range of EEG data signal bands. The sensitivity and precision of different bands also increase with the rate of 10%. The proposed algorithm compares with the Bayesian neural network (BNN) and support vector machine (SVM). Our experimental results validate that the proposed algorithm is better than BNN and SVM. All the experimental process performs on MATLAB14 software.

Motor Imagery, EEG Signals, Firefly Algorithm, ACO, Support Vector Machine, Classification