Multinomial Logistic Regression Model by Stepwise Method for Predicting Subjective Drowsiness Using Performance and Beha
The aim of this study was to explore the effectiveness of physiological and behavioral evaluation measures for predicting a drivers’ subjective drowsiness based on a multinomial logistic regression model. The participants were required to steer a steering
- PDF / 140,144 Bytes
- 10 Pages / 439.37 x 666.142 pts Page_size
- 42 Downloads / 223 Views
Abstract The aim of this study was to explore the effectiveness of physiological and behavioral evaluation measures for predicting a drivers’ subjective drowsiness based on a multinomial logistic regression model. The participants were required to steer a steering wheel and keep their vehicle to the centerline as much as they could, and to maintain the distance between their own car and a preceding car properly as much as possible using a brake or an accelerator. A number of measures were recorded during a simulated driving task, and the participants were required to report subjective drowsiness once every minute. EEG (electroencephalography), heart rate variability (RRV3), and blink frequency were the physiological measures recorded. Meanwhile, behavioral measures included neck bending angle (horizontal and vertical), back pressure, foot pressure, and tracking error in a driving simulator task. Drowsy states were predicted via a multinomial logistic regression model. Physiological and behavioral measures were independent variables in the regression model and equated to the dependent variable: subjective evaluation of drowsiness. The stepwise method was adopted for the estimation of parameters of multinomial logistic regression model. The interval used for attaining the highest prediction accuracy was a 100 s interval between 20 and 120 s before the prediction. This approach clarified that the parameters finally appeared in the multinomial logistic regression model were different among participants, which indicated that the optimal structure of the model for predicting subjective drowsiness should be different among participants.
A. Murata (&) Y. Ohta M. Moriwaka Department of Intelligent Mechanical Systems, Graduate School of Natural Science and Technology, Okayama University, 3-1-1, Tsushimanaka, Kita-ward, Okayama, Japan e-mail: [email protected] Y. Ohta e-mail: [email protected] M. Moriwaka e-mail: [email protected] © Springer International Publishing Switzerland 2016 R. Goonetilleke and W. Karwowski (eds.), Advances in Physical Ergonomics and Human Factors, Advances in Intelligent Systems and Computing 489, DOI 10.1007/978-3-319-41694-6_64
665
666
A. Murata et al.
Keywords Drowsiness Behavioral measure Multinomial regression model Stepwise method
Physiological measure
1 Introduction Bowman et al. [1] stated that approximately 30 % of crashes were caused by drowsiness. The development of a drowsiness detection system would be effective for the prevention of drowsy driving that becomes a potential risk factor of crashes. Understanding the gross tendency of reduced arousal level alone is insufficient to develop such a system. A more accurate prediction of the point in time when severe drowsiness occurs and potentially leads to higher risk of crash must be conducted. McDonald et al. [2] showed that the steering wheel angle was effective for predicting drowsiness-related lane-departures. Hanowski et al. [3] assessed drowsiness using an eye-closure measure a
Data Loading...
