Abstract:
The automation has made a great progress recently, but fully automation is still very difficult in complex task environment. Adaptive operation/support system becomes a viable solution. The real-time operator's mental workload (MWL) monitoring system is crucial for the design and development of adaptive operator-aiding/assistance systems. Although the data-driven approach has shown promising performance for MWL recognition, its major challenge lies in the difficulty of acquiring extensive labeled physiological signals. This paper investigates the ability of semi-supervised extreme learning machine (SS-ELM) to explore label samples and unlabeled samples. By comparing two different feature extraction methods, i.e., wavelet packet transform and the Hilbert-Huang transform, this paper tries to apply the SS-ELM algorithm to the challenging problem of operator's mental workload classification only by using a small number of labeled physiological data. Real data analysis results show that the SS-ELM method can effectively improve the accuracy and computational efficiency of the MWL pattern classification. Because the unlabeled training data can be collected from the operator's natural operations with less additional resources, a semi-supervised approach via unlabeled data can increase the efficiency of model development in terms of time and cost.