SCG Backpropagation Based Prediction of Stressed EEG Spectrum

In this paper, feature vectors pertaining to the changes in spectral transients of sleep EEG under hot environment has been studied using wavelet transforms and feed forward neural network is employed to detect the stressed patterns. Four continuous hours of sleep EEG recordings of subjects under the exposure of high temperature and also at room temperature have been filtered and visually analyzed. Three sleep stages: AWAKE, SWS (Slow Wave Sleep) and REM (Rapid Eye Movement) along with EMG and EOG activities under heat stress and without heat stress were quantified in time-frequency domain. Features extracted in terms of wavelet coefficients are further classified using scaled conjugate gradient algorithm (SCG). Classification accuracy of SCG algorithm is found to be 97.12% and 95.3% for stress and control subjects respectively, which may be considered as an efficient prediction model.