Two-stage pedestrian classification in automotive radar systems

Road and city traffic is dangerous for all car drivers and especially for pedestrians. Latest accident statistics [1] show that in 25% of all accidents pedestrians are involved. It is therefore important to protect especially those vulnerable traffic participants, e.g. by acoustic alarm signals or outside airbags, in case of an unavoidable accident. 24 GHz radar sensors with their all-weather capability are already in use in vehicles, and have great advantages compared with other automotive sensor systems. It has been described in [2], how the Doppler spectrum and the range profile are analyzed in a target recognition system for pedestrian detection. To increase the probability of correct classification, additional signal features have to be used. Also, a very fast target recognition system is required to respond and react quickly in dangerous situations. This paper is focused on a two-stage target classification system. Initially, the Doppler spectrum and range profile are extracted from the radar echo signal and applied in the recognition system. In a second step, additional features are calculated from the tracker which are fed back to the target recognition system to improve the system performance and to increase the probability of correct classification.