Classification of Stress Level using Approximate Entropy Features and Neural Networks

Most studies addressing the effect of chronic stress on health have reported that chronic stress is associated with an increased risk of infectious diseases [1] including HIV [2]. No significant indicator is available to measure the level of stress and it is one of the common research problems. In this paper, to recognize the level of stress, a simple stress level classification system has been proposed using brain wave electroencephalogram (EEG) signals. The proposed stress level classification system records the brain wave signals while listening to the sound clips mixed with noise. The recorded Electroencephalography signals are pre-processed and segmented into four frequency bands. The band frequency signals are used to extract features using band energy and approximate entropy (Apen) algorithm. The extracted features were then associated to the stress levels evaluated by subjective evaluation test. A simple multilayer neural network model has been developed to classify the level of stress. The proposed methods are validated through simulation.