Detection of mental stress using novel spatio-temporal distribution of brain activations

Detection of mental stress is an important research problem as it is essential for ensuring overall well-being of an individual. Recently, various physiological sensor signals are being used by researchers for the said purpose. In this study, we have used electrical brainwaves recorded using EEG device for detection of mental stress. Literature suggest that due to volume conduction effect of skull and underlying tissues, the recorded EEG signal is not a true representative of the actual changes in the brain responses that we intend to measure. To address this issue, instead of working with the direct features computed from the recorded signals, we have relied on spatio-temporal transition behavior of those features. It helps to reduce the limitations posed by the low cost devices. Towards that aim, we have used publicly available DASPS dataset containing EEG recordings of 23 participants while engaged in an exposure therapy (where a stressful situation was narrated by a trained psychologist followed by recall of situation by participants). We elaborated the process of calculating spatio-temporal transition of features in detail. Results suggest that our proposed spatio-temporal transition based feature is able to discriminate various levels of mental stress of an individual. It is observed that the maximum classification accuracy obtained is 83.8% for both binary and 4 class classifications. We compared the performance of our proposed approach with state-of-the-art results and observed that our proposed approach with proposed spatio-temporal transition based features performs well for classification of mental stress and outperforms approaches reported in literature.