Classification of Boredom and Anxiety in Wrist Pulse Signals using Statistical Features

In Traditional Indian Medicine (TIM), the characteristics of Wrist Pulse Signals (WPS) are used as an indicator of health status. In this work, the study has been carried out to show the competency of WPS to detect human emotions during Boredom and Anxiety states. The whole setup includes WPS amplification, filtration, data acquisition and stimulus program to induce definite state of mind (emotion). ®LABView software has been used for acquisition and analysis of pulse signals. Statistical parameters were derived for classification of emotional states. The significance of these parameters has been tested using the t-test and p-values. Further, comparison of Naive Bayes, SVM, MetaMulti class and Random Forest has been shown for 3 different cases: Boredom vs. Neutral, Anxiety vs. Neutral and Boredom vs. Anxiety. Among all the classifiers, SVM showed the best results with an accuracy of 75% and 62.75% in detecting Anxiety and Boredom states respectively. Comparative analysis has been carried out for WPS under different states demonstrating the effect of various emotions (feelings) changes on WPS. The experimental outcome demonstrated a noteworthy improvement in classifying boredom and Anxiety states using WPS.