Mental Stress Detection Based on Soft Computing Techniques

In this study, a novel approach is proposed for mental stress recognition through automatic analysis of eye video sequences. The proposed system consists of five stages: video capturing, fuzzy image processing, signal processing, feature extraction and, classification. The pupil parameters including Pupil Diameter (PD) and Pupil Dilation Acceleration (PDA) are measured using soft computing techniques wherein the eye region is detected using the genetic algorithm (GA), and a fuzzy filter is designed for noise reduction. Edge detection is performed based on fuzzy reasoning and linking is done using Hough transform. Then, signal processing technique is applied to the pupil parameters to extract their most relevant features. Extracted features are imported into the learning system to classify the affective states between "stress" and "relaxed". The Fuzzy SVM (FSVM) is applied to this classification process. In order to induce the stress in subjects, a Stroop color-word test is designed. Also, the results obtained from the pupil parameters are compared with two other physiological signals including Electrocardiogram (ECG) and Photoplethysmogram (PPG). The experimental results indicate the pupil parameters have great potential for stress recognition compared to the other two physiological signals and, proposed stress recognition system is promising.