Edinburgh Napier University

  Stress analysis by utilizing electroencephalography (EEG) device in conjunction with signal processing techniques has emerged as an important area of research and the efforts are being made on detecting and classifying stress level. Non-invasive EEG device is used in this study to collect brain signals and analyze the signals by applying the modified Welch’s fast Fourier transform (FFT) algorithm to extract the power spectral density (PSD) of each frequency band and calculate the power ratio of Alpha to Beta and Theta to Beta. The analysis of the power ratio has further validated that the Theta/Beta power ratio can be used as feature of stress and thus, imported its dataset into k-means clustering to divide the subjects into three categories. Lastly, the clustering model is fed into support vector machine (SVM) to classify three-level stress which are of low, moderate and high. The result has signified the feasibility and effectiveness of the three-level stress classification at overall classification accuracy of 90% by applying Theta/Beta power ratio of the brain signals as well as using SVM classifier. The outcome of the research suggests the proposed method can be used for the implementation of stress monitoring system.