Dynamics of Blink and Non-Blink Cyclicity for Affective Assessment: A Case Study for Stress Identification
Peng Ren,Armando Barreto,Xiaole Ma,Shengnan Liu,Min Zhang,Ying Wang,Yeyun Dong,Dezhong Yao
DOI: https://doi.org/10.1109/taffc.2019.2946829
IF: 13.99
2022-01-01
IEEE Transactions on Affective Computing
Abstract:Previous studies have shown that eye activities, including blinks, can indicate the psychological state of an individual. However, almost all previous studies analyzing blinks merely concentrated on traditional descriptive statistics, which are unable to reflect their dynamic processes. Furthermore, the states of non-blink (opening the eyes) and blink alternate with each other, forming a physiological cycle. If we only investigate blinks alone, it may be inadequate to describe how blinking works. Therefore, we attempted to recognize the affective state ("relaxation" vs. "stress") of an individual through the dynamics of blink and non-blink cyclicity (BNBC), as one example, to illustrate this method. First, the "Stroop Test" was employed for emotion elicitation. Then, features were extracted from a categorical time series (0: non-blink; 1: blink), which was recorded by the eye-tracking system. Finally, the areas under the receiver operating characteristic curve (AUC) values were obtained via eight commonly used classifiers. The results show that, compared with the traditional approaches for blink analysis, BNBC exhibits more compelling proficiency to detect stress. In summation, BNBC can be considered a new type of psychophysiological measure, which could be widely applied in psychology, medicine, and engineering.