This letter proposes a method to detect the angular locations of human targets using a multiple-input-multiple-output (MIMO) monopulse radar system. Sum and difference beams used for monopulse processing are simply implemented by digital beamforming of a virtual array generated by collocated MIMO radar. Human targets are readily distinguished from stationary clutter based on the phase variations in their signals, which are detected by estimating the curve lengths (CLs) of the I/Q trajectories. A monopulse processing technique integrated with the CL method is thus proposed. This system resolves two human targets, of which the angular separation is less than the angular resolution of the conventional radar. The system is verified through both simulations and experiments in an indoor environment.