Ultra-wideband (UWB) radar imaging has attracted attention for use in security and intelligent transportation system (ITS) applications. Conventional UWB Doppler interferometry is an effective way to obtain high-resolution images while using a simple radar system. However, this method produces ghost images when multiple closely-spaced human targets are present. To resolve this problem, we propose a new technique that combines UWB Doppler interferometry with an adaptive beamforming method called estimation of signal parameters via rotational invariance techniques (ESPRIT). We also propose a tracking and separation algorithm that uses the k-nearest neighbor method. Through a combination of numerical simulations and measurements, we demonstrate the remarkable performance improvement that can be achieved using our proposed method. The proposed method can separate multiple humans with a root-mean-square error of 5.2 cm, which makes its accuracy 1.9 times higher than that of the conventional method.