Estimation Accuracy of Driver’s Mental Workload on Difference in Number of Data Samples on Eye Movements

Abstract

There are several studies that discriminant analysis and machine learning using parameters related to eye movements and head movements apply to estimate the driver’s mental workload while driving. However, all of the eye movement measuring devices used in these experiments have extremely high time resolution, which is not desirable in terms of cost when developing a system in consideration of practical use. Therefore, the purpose of this experiment was to clarify the minimum value of data measurement sampling that enables extraction of eye movement parameters that can evaluate the difference in mental workload due to the driver’s N-back task. Therefore, we investigated the relation between the estimation accuracy of the driver’s mental workload and the sampling rate when measuring eye movements. As a result, it was shown that the possibility of estimating the driver’s mental workload using eye movement parameters even when the eye movement sampling rate is low.