Target/clutter disentanglement using deep adversarial training on micro-Doppler signatures
暂无分享,去创建一个
In this manuscript we argue that, under certain conditions, machine learning techniques can help to increase the signal to background level of a target signal to aid the detection and classification process in the radar signal processing chain. Specifically, the deep adversarial training concept, through the use of Denoising Adversarial Autoencoders (DAEs), has been applied for the problem of separation the micro-Doppler signatures of wind-turbine and drones, in order to be able to extract the latter for further detection and classification purposes.
[1] J. J. M. de Wit,et al. Classification of small UAVs and birds by micro-Doppler signatures , 2013, 2013 European Radar Conference.
[2] Soumith Chintala,et al. Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks , 2015, ICLR.
[3] G. P. Cabic,et al. Radar micro-Doppler feature extraction using the spectrogram and the cepstrogram , 2014, 2014 11th European Radar Conference.
[4] Yoshua Bengio,et al. Extracting and composing robust features with denoising autoencoders , 2008, ICML '08.