Detecting the early drop of attention using EEG signal

Abstract

The capability to detect the drop of attention as early as possible has many practical applications including for the development of the early warning system for those who involve in high-risk works that require a constant level of concentration. This study intends to develop such the capability on the basis of the data of the brain waves: delta, theta, alpha, beta, and gamma. For the purpose, a number of participants are asked to participate in the study where their brain waves are recorded by using a low-cost Neurosky Mindwave EEG sensor. In the process, the participants are performing a continuous performance test from which their attention levels are directly measured in the form of the response time in conjunction to those waves. When the response time is much longer than a normal one, the participant attention is assumed to be dropped. A simple k-NN classification method is used with the k = 3. The results are the following. The best detection of the attention drop is achieved when the attention features are extracted from the earliest stage of the brain wave signals. The brain wave signal should be recorded longer than 1 s since the time the stimulus is presented as a short signal leads to a poor categorization. A significant drop in the level of response time is required to provide the brain signal that better predicts the change of the attention.