Research On Detection And Analysis Of Mental Workload Evoked By N-back

Nowadays, with the development of technology, working tasks have becomemore and more intellective. This change brings about much mental workload inoperators. Therefore, mental workload is becoming an increasingly important issue intoday’s society. Due to the variety of tasks, there exist lots of difficulties for practicalapplications. One of the difficulties associated with mental workload is themultidimensional nature of the tasks, so it is important to figure out the differencebetween tasks.In order to investigate the similar technology of mental workload detection fordifferent tasks, this study designed two kinds of n-back tasks. They were Verbaln-back and Spatial n-back with four varied memory load levels (0,1,2, and3-back)respectively. Electroencephalogy (EEG) signals of17healthy volunteers wereacquired during the experiment and then analyzed using AR model, event-relateddesynchronization/event-related synchronization (ERD/ERS), sample entropy andwavelet multiscale entropy to investigate different characteristics between tasks. Inaddition, support vector machine (SVM) was utilized to classify four levels of mentalworkload. The results showed thatthere existed significant differences between ARmodel and ERD/ERS among different tasks. Additionally, the classification resultsindicated that the recognition rates of AR features and combined features were thehighest which could reached98%.Furthermore, in order to improve the performance, recursive feature eliminationbased on support vector machine (SVM-REF) was adopted. The optimized methodnot only increased the recognition rates but also decreased the number of channels.These measures were proved to be sensitive indicators of mental workload and coulddifferentiate two tasks in four levels. The research could provide a technical supportfor the practical application in the detection of mental workload, improving theaccuracy and convenience.