Monitoring driver's attentiveness is crucial for transportation safety. In this article, inattentive driving behavior detection based on frequency-modulated continuous-wave (FMCW) radar systems is proposed for this purpose. Seven typical driving behaviors which result in reduced attentiveness are involved in this study. Time-Doppler spectrogram and range-Doppler trajectory are utilized to analyze their features in multiple domains, including time, Doppler, range, and radar cross-section (RCS). These features are extracted as inputs to a machine learning classifier to obtain recognition results. Extensive experiments on a real car environment have been conducted to show its feasibility and superiority by obtaining an average accuracy rate of around 95%. The influences of radar center frequency, individual diversity, and radar view angle are also investigated.
- Driving behavior monitoring
- frequency-modulated continuous-wave (FMCW) radar
- machine learning