In ‘radiation remote sensing’ multiple unknown high energy sources are generally involved. The detectors, upon sensing the corresponding mixed signals, must separate their contributions blindly for further analysis. A practical way to perform this separation could be through the Independent Component Analysis algorithm. However, the challenge faced is that theoretically there is no correlation among events, even those arising from the same source – thereby disabling meaningful ICA analysis. We overcome this hurdle by use of a thin barrier and by providing wide detector pulses. The radiation events that interact with the barrier take a longer time to reach the detector due to their increased path length. They also lose some energy, which makes them increasingly prone to capture in the barrier once they have scattered. These observations are confirmed through Monte-Carlo simulations upon Gamma-ray sources. Normalized crosscovariance up to 0.22 was found, but is actually controllable through appropriate selection of the detector shaping-pulse width. Experiments on a physical setup confirm these findings. Finally, the application of the ICA approach is demonstrated to demix, or separate, the individual contributions of the sources to the observed detector signals.