This paper proposes a novel human-aware energy beamforming (BF) method for safe and secure microwave wireless power transfer (MWPT). The proposed method first estimates an equivalent channel between a transmitting array antenna and a human body from continuous-wave pilot signals Doppler-shifted by human vital activities by utilizing the MWPT system as a quasi-monostatic multiple-input multiple-output (or single-output) radar. This allows the online estimation of the radio frequency (RF) exposure power without prior human geometry information. Then, the closed-form optimum BF weight can be derived to maximize the figure of merit (FoM), defined as the ratio of the received power to exposure power in the form of a generalized Rayleigh quotient. This simultaneously facilitates the suppression of RF exposure and enhancement of transfer efficiency. Comprehensive experiments were conducted to evaluate the proposed BF performance using our multi-antenna testbed with proven applicability to various human-sensing experiments. The indoor experiment proved that the proposed method performs null steering auto-tuned to human positions within a distance of up to 3~m and to the physical characteristics of six subjects, even under realistic conditions with multipath fading and array error. It achieved the highest FoM of 21.65 dB among conventional BF methods for a typical case. This was an improvement of 7.10 dB when compared with the result of maximum ratio transmission BF, which maximizes the transfer efficiency. This study demonstrates the possibility of coexistence of humanity and MWPT technology for its future social implementation.