A Method for Classifying Cognitive Load of Visual Tasks Based on Eye Tracking Features

The measurement of the cognitive load has always been a fundamental technology in human-computer interaction and human factors because cognitive overload will not only cause the reduction of work efficiency and subjective satisfaction of human-computer interaction but also may affect working safety. In this paper, we propose a method to classify the state of cognitive load by collecting eye movement data and extracting features in time domain and frequency domain. We designed a complete experimental procedure, including reading comprehension, contrast judgment, and content search, and collected four classes of experimental data under different difficulties. Some features of eye movement were selected by the ReliefF feature ranking method. Finally, we trained five classifiers to classify the eye movement data of different cognitive load levels. The experimental results show that the highest accuracy of the four classes can reach 90.6%.