A self-aligning artificial knee mechanism with two degrees of freedom was presented aiming at the problem of human-robot system misalignment at the knee joint. The mechanism was driven by two motors to simulate the flexion and extension movements of the biological knee joint. A kinematic model of human-robot coupling was established to quantify the misalignment as the deviation of the swing angle and position of the mechanism’s connecting point from the reference point of the calf. A particle swarm optimization (PSO)-based end-matching method for human-robot systems was proposed in order to reduce deviations and optimize the parameter of the key component. The swing angle of the artificial knee joint approaches the knee flexion and extension angle in the process of the bandage point approaching the reference point. A group of high-speed cameras was used to measure the movement data of individual lower leg, and the human-robot matching operation was conducted with the results as a reference. A digital virtual prototype was developed to conduct simulation experiments. The matching effect of the artificial knee joint and individual calf motion was verified through a joint simulation of MATLAB-Adams. Results show that the artificial knee joint can achieve motion matching and self-alignment of the swing angle.