DWT-based feature extraction and classification for motor imaginary EEG signals

A brain-computer interface (BCI) permits cerebral activity alone to control the external devices for assisting people with neuro muscular impairments. Electroencephalogram (EEG) signals are used for brain computer interaction which is highly non-stationary therefore major challenge is to extract features and classify the signals accurately. In this paper we focused on the extraction of features of EEG motor activities using Discrete Wavelet Transform (DWT) and classified the signal for using Artificial Neural Network (ANN) for differentiating left and right hand imagery movement. Two sets of feature vectors are taken from beta rhythm as input to the Feed-forward neural network classifier. We observed that three input feature vectors like mean, standard deviation and peak power achieved better classification performance result of 80.71% compared to two input feature vector which is of 78.57%.