Prediction And Detection In Change Of Cognitive Load For VIP's By A Machine Learning Approach

Abstract

The significance and urgency of detecting the cognitive load of a Visually Impaired Person (VIP) are essential when perception comes while designing an automated navigation aid for them in unfamiliar indoor environments. Our paper presents a novel and robust framework based on the iterative feature pooling technique which recursively selects paramount features that maintains relationships with the change in the cognitive load of the brain. We took the well-established Event-Related Desynchronization and Synchronization (ERDS) method for indexing the cognitive load and further developed the work by operating with the band power of not only the Alpha wave but the Alpha Beta band power ratio and Alpha Theta band power ratio. The supervised machine learning classifier, Gradient Boost outperformed all other classifiers reaching 94% accuracy in the best case. When provided with the most reliable features and proper tuning, this turns out to perform 7% to 8% better than the other classifiers like the Support Vector Machine (SVM), K-nearest neighbor, Naive Bayes, and Multilayer perceptron. We considered some performance parameters like the accuracy, null-accuracy, recall, precision, F1 Score, and False Alarm rate to evaluate the performance of all available supervised Machine learning classifiers. Our paper marks out the estimation of cognitive load based on Electroencephalogram (EEG) signals analysis with the existing literature, background, leeway, features, and machine learning techniques.