METHOD, APPARATUS AND COMPUTER PROGRAM FOR CLASSIFYING RADAR DATA FROM A SCENE, FOR TRAINING ONE OR MORE NEURAL NETWORKS FOR CLASSIFYING RADAR DATA

The present disclosure relates to methods, apparatuses and computer programs for classifying radar data from a scene and for training one or more neural networks to classify radar data. The method comprises the following steps: acquiring radar data from a scene and determining rhythm speed data and micro-distance Doppler data from the radar data; encoding the rhythm velocity data and the micro-distance Doppler data using the first and second trained autoencoders to obtain a rhythm velocity feature vector and a distance Doppler feature vector; decoding the rhythm velocity feature vector and the distance Doppler feature vector using first and second trained decoders to obtain reconstructed rhythm velocity data and reconstructed distance Doppler data; determining first reconstruction loss information based on the rhythm velocity data and the reconstructed rhythm velocity data, and determining second reconstruction loss information based on the distance Doppler data and the reconstructed distance Doppler data; th