Analyzing the classification capability of Micro-Doppler spectra

The Micro-Doppler spectra of a frequency modulated continuous wave (FMCW) radar system can be used to identify objects in a surveillance application scenario. We investigated different classification approaches to determine their potential to classify different objects. A radar sensor in the 24 GHz band was used to record scenes with three different moving objects. We compared classification methods which recognized fairly short segments of the recorded scene to other approaches which analyzed the whole sequence of Doppler spectra with a Gaussian Mixture Hidden Markov modelling (GMM-HMM) and with a long short-term memory (LSTM) based neural network. High recognition rates of up to 99% could be achieved on the recorded test data set.