2021/08/23 by Fred Atilla, Atilla, Fred, Maryam Alimardani +1
Medicine · Neuroscience · Psychology · #Artificial Intelligence (cs.AI) #Cognitive Functions and Memory #ECG Monitoring and Analysis #EEG and Brain-Computer Interfaces #FOS: Computer and information sciences #Human-Computer Interaction (cs.HC)
paper · pdf · doi:10.48550/arxiv.2108.10062
openalex publication_date 2021/08/23 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Accurate detection of a drivers attention state can help develop assistive\ntechnologies that respond to unexpected hazards in real time and therefore\nimprove road safety. This study compares the performance of several attention\nclassifiers trained on participants brain activity. Participants performed a\ndriving task in an immersive simulator where the car randomly deviated from the\ncruising lane. They had to correct the deviation and their response time was\nconsidered as an indicator of attention level. Participants repeated the task\nin two sessions; in one session they received kinesthetic feedback and in\nanother session no feedback. Using their EEG signals, we trained three\nattention classifiers; a support vector machine (SVM) using EEG spectral band\npowers, and a Convolutional Neural Network (CNN) using either spectral features\nor the raw EEG data. Our results indicated that the CNN model trained on raw\nEEG data obtained under kinesthetic feedback achieved the highest accuracy\n(89%). While using a participants own brain activity to train the model\nresulted in the best performances, inter-subject transfer learning still\nperformed high (75%), showing promise for calibration-free Brain-Computer\nInterface (BCI) systems. Our findings show that CNN and raw EEG signals can be\nemployed for effective training of a passive BCI for real-time attention\nclassification.\n