Enhanced operator fatigue detection method based on computer-keyboard typing style

摘要

Many operators are working in jobs that require stressful mental tasks such as transportation supervision, vehicle driving, banking and others. Prevention of fatigued-based human error, that has been a standing challenge in such work areas, can be detected and quantified using human performance level. This paper proposes an enhanced method for operator fatigue detection based on computer-keyboard typing style. This is achieved by using an existing dataset for psychomotor impairment detection using natural typing style on computer keyboard. Several fatigue-related features are extracted and fed to two parallel classifiers based on artificial neural network (ANN) and support vector machine (SVM) algorithms. Outputs of those classifiers are then combined to enhance the classification performance, using Bayesian combiner. Performance of the developed fatigue detection system is assessed experimentally in terms of the classification accuracy as compared to a ground-truth dataset. The obtained results demonstrated that utilization of the Bayesian combiner has significantly improved the fatigue detection accuracy (94%) as compared to that of the ANN (87.5%) and SVM (91%) classifiers. These findings are favorably compared to the state of the art but with easily identified fatigue-related features.