Effect of Sliding Window Time on the Classification of Driver Mental Workload Performance Using Near-Infrared Spectroscopy (NIRS)

Near-infrared spectroscopy (NIRS) is a promising solution for classifying mental workload while driving, which is important for creating a support system and preventing traffic accidents. Most of the existing research classified mental workload levels via machine learning or deep learning algorithms, where the sliding window time plays an important role in obtaining the features. This study investigated a novel classification of mental workload level using NIRS data by the difference of sliding window time in both channel-independent and subject-independent scenarios. Five sliding times from 1 to 5 s were tested for classification of the mental workload level in 60 experiments in the naturalistic condition. The results indicated that the sliding times were not significantly related to the accuracy of classification using NIRS data when the sample size was just enough. Furthermore, this study succeeded in indicating the driver mental workload using NIRS information in the naturalistic situation with autocross and car-following conditions with an accuracy of over 80%.