WAVELET TRANSFORM AS A PREPROCESSOR FOR MENTAL TASK CLASSIFICATION USING EEG SIGNALS

摘要

The paper presents an artificial neural network (ANN) based approach for classifying mental tasks using electroencephalogram (EEG) signals. The wavelet transform is used as a preprocessor to reduce noise associated with EEG signals due to power line interference and movements of the electrodes on the scalp during EEG recordings. Linear predictive coding (LPC) technique is utilized to extract features representing the processed EEG signals. Extracted features are then applied as inputs to a NN based classifier. A postprocessor module is added to cancel the classification errors due to artifacts corrupting EEG signals. Five mental tasks are considered. The implemented system classifies all possible pairs of the five tasks with an average classification accuracy of 96.1%. It also achieves an average classification accuracy of 91.2% for five mental tasks classification.