Distracted Driving Performance Measures

Abstract

This study introduced and applied the signal processing algorithm fast Fourier transform (FFT) to the calculation of vehicle-control variations during distracted driving. This methodology has proved to be useful in extracting behavioral features by converting a signal from the time domain to the frequency domain. Previous studies generally used the variations calculated in the time domain, such as standard deviation, as an indicator to predict distracted driving. In this study, the use of FFT and spectral power analysis provided new prospective ways to analyze driver performance and vehicle-control data. The spectral power analysis showed that cell phone use resulted in vehicle lateral control variations. Drivers' lane position–keeping profiles were bumpiest during visual–manual tasks, as demonstrated by the largest average spectral power value and the greatest variation range compared with the other two conditions. Baseline driving appeared to have the smoothest lateral controls. Older drivers had the highest lateral control variations among the three age groups when conducting visual–manual tasks; this result suggests that they are less capable of controlling the wheels while engaging in secondary tasks that require their visual and manual inputs. The results of this study suggest that the FFT method provides a meaningful measure in extracting variations in distracted-driving behavior. Comparisons between the features in the frequency domain and the time domain should be conducted in future studies.