Measure The Level Of Cognitive Load Based On Eye Movement Data

Cognitive load refers to the cognitive resources consumed in the process of information processing.Due to the limited cognitive resources,the level of cognitive load will affect the work efficiency.A scientific method to measure cognitive load is widely needed in the many areas,such as improving educational methods,designing interactive products,and monitoring high-pressure jobs.There are three kinds of cognitive load measurement methods,which are subjective measurement,task performance measurement,and physiological measurement.The method of subjective measurement uses questionnaires to investigate people's subjective feelings to measure cognitive load,which is easily affected by subjective factors.Task performance measurement uses the performance of the task performer to assess cognitive load.Though the performance is easily quantified,parameters for evaluating the performance must be specifically set according to the task.Physiological measurement is objective and quantifiable,and eye movement data can be collected in a non-contact manner,so this method has high practical value.This thesis designs cognitive load eye movement data collection experiment according to the theory of cognitive load,analyzes the features of eye movement regarding to cognitive load,proposes a method for eliminating individual difference in features,and measures the cognitive load based on the theory of pattern recognition.This study is divided into two phases,and the main research contents are as follows:In the first stage,two kinds of cognitive load states are identified.The judgment task is used to induce cognitive load.According to statistical tests,12 features that could reflect the state of cognitive load are identified.A method used to eliminate the feature's differences between subjects is proposed.Support Vector Machine(SVM)is used to recognize the cognitive load states,and the recognition accuracy is 90.2%.The optimal combination of features for identifying the states of cognitive load are determined according to the recognition results.In the second stage,the quantification of cognitive load is realized.The level of cognitive load is controlled by question difficulty in mental arithmetic task;55 eye movements are extracted,the level of cognitive load and the dwell time ratio of interest area and the change of pupil size during calculation are analyzed.Sequential Backward Selection(SBS)and SVM are used to determine the optimal combination of features and to recognize the multiple levels of cognitive load,the recognition accuracy is 74.4%.The cognitive load is quantified by using the posterior probability of classificationThis thesis provides a method to identify cognitive load states and levels by using eye movement data,and complete quantification the level of cognitive load.The thesis illustrates the feasibility of measuring cognitive load levels with eye movement data.Moreover,this method can acquire data in a non-contact,which makes the method practical in real application.