In this paper, we present the design and evaluation of a novel control method for a power assist robotic system for manipulating objects with it by human subjects. We consider the manipulation of objects in harmonic motion (object is lifted up and lowered down repeatedly) as we think that this motion may be very practical and useful in object manipulation. We develop a 1 DOF power assist system. The subjects manipulate different sizes of objects with the system in harmonic motion. We analyze human characteristics such as weight perception, load force, object motions etc. We find that the load force and acceleration are excessive that reduce the performances of the system. We then optimize the perceived heaviness. We then design and implement a novel control method based on human characteristics to reduce load force and acceleration. Results show that the novel control reduces load force and acceleration, optimizes perceived heaviness and thus makes the performances (maneuverability, safety etc.) satisfactory. We also compare this novel control designed for object manipulation in harmonic motion to another novel control designed for lifting objects with power-assist in linear vertical motion. Finally, we propose to use the findings to develop control method for power assist devices for manipulating heavy objects in industries that may enhance interactions between humans and robots.