Classification of Vulnerable Road Users Based on Spectrogram Autocorrelation Features

In the upcoming years on the path towards a climate-neutral economy, the urban traffic system plays a decisive role in the European Union. Due to the growth of bike-sharing and overall amount of e-bike users, the number of cyclists as well as pedestrians will raise. The increase of vulnerable road users in urban city environments along with the recent development towards automated and autonomous driving sets the demand of more sophisticated sensor and classification systems. This paper presents a method of classifying real radar measurements of pedestrians and cyclists measured with a 77 GHz continuous-wave Doppler radar. The approach shows that the autocorrelation of the time-frequency domain exhibits significant features that can be used to differentiate between the classes as well as variations within the motions.