Classification and Identification of Micro-Doppler Signatures of Pedestrian and Cyclists Using Simulated Data and ELM

There are great interests to study detection, and classification of pedestrian and cyclists using low-cost frequency modulated continuous wave (FMCW) radar sensors for intelligent transport systems (ITS). This paper presents a study about detection and classification of pedestrian and/or cyclists based on realistically simulated FMCW radar sensor echoes. From radar echoes, the micro-Doppler signatures of different targets can be observed through time-velocity diagrams and then extracted. Using a relatively small number of data samples for training, an Extreme Learning Machine (ELM) can be trained to detect the number of targets and classify their types. Numerical experimental results show the effectiveness of the proposed approach and the potential of low-cost radar sensors in ITS applications.