Human Cognitive Control Mode Estimation Using JINS MEME

Abstract

In this study, a human cognitive control mode to determine human reliability is estimated using data obtained by eyewear called JINS MEME. This device can measure electro-oculography, acceleration, and angular rate without load on the subject. The computer game widely known as TETRIS is selected as a test task because the task difficulty affecting the subjects’ workload can be controlled by varying game speed. NASA-TLX is utilized to measure the subjective mental workload of 12 subjects. First, the blink rate measured by JINS MEME focusses as a parameter to reflect mental workload. Consequently, 10 out of 12 subjects show negative correlation between blink rate and NASA-TLX score. Thus, basic validity using this device to estimate mental workload is confirmed. Second, a machine algorithm, i.e., a support vector machine (SVM), is applied to the measured data to categorize the subject’s conditions into either Normal or Degraded. Additional parameters measured by JINS MEME are introduced as input data to the SVM. It is demonstrated that the SVM can categorize the subjects’ conditions with 80-90% accuracy, and that the reduction of input variables using principle component analysis results in higher categorization accuracy.