This thesis contributes to the field of short-range ultra-wideband (UWB) Radar. In particular, an object recognition approach performed by a bi-static UWB Radar has been investigated in this thesis. The investigated objects consist of simple canonical and some polygonal complex objects which are scanned on a circular track at about 1 m distance. Geometrical features, texture features and moment based features are extracted from the Radar data to carry out the recognition. Yet, the precise temporal evolution is subject to massive distortions, mainly caused by severe interference conditions and transient effects of the hardware. Thus, super-resolution algorithms have been developed which go far beyond the classical bandwidth given resolution and asked for research on various fields: (i) An innovative wavefront extraction algorithm with polarimetric diversity exploitation has been developed to separate pulses which overlap almost the whole pulse duration; (ii) a highly precise feature extraction algorithm has been developed which localises significant scattering centres by processing the previously extracted wavefronts; (iii) a novel UWB object recognition algorithm has been developed to classify and discriminate the resulting microwave images. When scanning objects from all sides, exceptional recognition of objects was achieved by a minimum mean squared error classifier. Further improvement in recognition was obtained, especially at severly restricted tracks, by the application of Bayes theory which constitutes a superior classifier to the above. In addition to the main field of research, a novel stereoscopic 3D UWB imaging algorithm, based on a spatially spanned synthetic aperture in conjunction with ellipsoidal shaped wavefronts, has been developed. The ultimate test of any model and system is an experimental validation. Consequently in this thesis, all developed algorithms and the object recognition as a whole system are experimentally validated within an elaborate measurement campaign.