Stress Level Detection Using Hybrid Features and Meta Model Classifier with EEG Signal

Abstract

Stress detection has become a critical area of study due to its significant impact on human health and productivity. This paper presents a novel approach to stress detection that leverages a hybrid Features  and a hybrid classifier framework, integrating various machine learning techniques to enhance accuracy and reliability. Our hybrid features combines selected time domain & Frequency domain features for selected channel to improve the accuracy of the result. The hybrid classifier framework employs an ensemble of machine learning algorithms, including Support Vector Machines (SVM), Random Forests, and XGBoost, to improve classification performance through diversified model strengths. The proposed methodology was evaluated using benchmark datasets. Our results demonstrate that the hybrid approach significantly outperforms traditional single-source and single-classifier models, achieving higher accuracy, precision, and recall in stress detection. Using the proposed hybrid techniques, we achieved a higher accuracy  by considering  base classifiers and the meta-classifier.