Prediction of drowsiness using multivariate analysis of biological information and driving performance

The aim of this study was to predict drowsy states by applying multivariate analysis such as discrimination analysis and logistic regression model to biological information and establish a method to properly warn drivers of drowsy state. EEG, heart rate variability, EOG, and tracking error were used as evaluation measures of drowsiness. The drowsy states were predicted by applying discrimination analysis and logistic regression to these evaluation measures. The percentage correct prediction for discrimination analysis and logistic regression were 85% and 93%, respectively. The polynominal logistic regression model was found to lead to higher prediction accuracy. The biological data might be used for the long-term prediction of drowsiness, while the performance data such as tracking error can be used only for the short-term prediction.