Identification of alertness state based on EEG signal using wavelet extraction and neural networks

Abstract

This research proposed identification of alertness state using wavelet transformation and neural networks. Wavelet extracted three wave components of an electroencephalogram (EEG) signal, namely alpha, beta and theta as input of neural networks, so that reduce 256 Hz recording into 28 data each second. It also used a asymmetric of the channel that improved recognition. EEG signals was obtained from four subject as training data which was tested to other four subjects. Each subject recorded with blue light stimulation during eight minutes and continued no stimulation in last eight minutes. It was sit position. Light stimulation was intended to increase alertness the subject. This research focused on develop identification of alertness state system. Based on the results of testing the system on the new data showed that C3 and C4 channel made best classification with 81% recognition accuracy. From all data, less alert state gave recognition more than other. Characteristic of data was significant factor of the result.