In this paper, a simple method to reduce the number of EEG channels for a Motor Imagery based Brain Machine Interfaced (BMI) system has been proposed. By reducing the number of EEG channels, the number of features can be reduced and this has to be achieved without sacrificing the classification accuracy and computational time of the BMI. EEG signal are recorded from ten subjects using a 19 channel EEG amplifier. Spectral Energy Entropy feature are extracted from the recorded signal. The extracted features were then used to model a neural network. A simple statistical analysis based on standard deviation and kurtoses of the extracted features were then used for the channel reduction process. The classification accuracy of the neural network model formulated with the 19 channels features were compared with the classification accuracy of the model with the features of the channel selected using statistical approach. From the results it is observed that using statistical methods, the number of features can be reduced without sacrificing its classification performance.