Real time estimation of task specific self-confidence level based on brain signals

Abstract

Self-confidence is one’s own belief of success with respect to a specific task. Cognitive tasks like decision making, problem solving etc. are influenced a lot by our self-confidence level. Recent studies show possibility of assessing self-confidence level from brain activation patterns. Here, we are proposing a novel metric based on brain signals acquired by Electroencephalogram (EEG) for quantification of self-confidence level of an individual. The EEG signal is recorded using a low cost, low resolution EEG device. Most significant EEG features (with p value <0.05) have been identified and used for formulating a metric for self-confidence level. Training data is refined using Random Sample Consensus (RANSAC) method to remove the outliers. With these stable training data, reference clusters for low and high confidence categories have been formed based on the most significant features. These reference clusters are used thereafter to compute the confidence level quantitatively for the test participants. Results show that the proposed metric can classify low and high self-confidence levels satisfactorily. Thus, the proposed metric provides a quantified measure of self-confidence level with which a task is performed. The metric can also be used in various applications like confidence measurement in academic settings, online tests, cognitive assessment, rehabilitation, therapy design.