Currently, using micro-Doppler information extracted from radar signals to estimate human breathing has become a promising solution in health-care and fire rescue. However, accurate extraction and distinction of breathing information from multiple human targets is still a difficult problem to be solved. In this paper, we propose an effective algorithm for detection of human breathing in multi-person scenarios. The proposed algorithm separates radar signal by estimating the range and angle of arrival of different human objects, and then estimates the frequency of breathing by extracting micro-Doppler features frame by frame. In this way, we can get both the position and corresponding breathing frequency of each human object in multi-person scene. Numerical simulation and real measurements with a self-designed 24GHz MIMO-FMCW radar are both provided for validating the proposed algorithm.