Instantaneous Mental Workload Classification Using Semi-Supervised Learning

Abstract

The accurate monitoring and analysis of operator’s mental workload (MWL) is a critical component for design and implementation of adaptive/intelligent human-machine collaborative systems in various safety/mission-critical application areas. Although data-driven approach has shown potential for MWL recognition problem, in many practical applications (such as mental workload assessment using physiological data under the present investigation) it is usually difficult or expensive to acquire a sufficient number of labeled data for training the data-driven machine learning models. This paper develops a semi-supervised extreme learning machine (SS-ELM) for the MWL pattern classification problem based on a small number of labeled data and a large number of unlabeled data. The experimental data analysis results have shown the effectiveness of the proposed SS- ELM paradigm for improving the accuracy and computational efficiency of MWL classification. The proposed semi-supervised learning paradigm may provide an alternative machine learning approach to utilizing the large number of unlabeled data collected under realistic operational environments.