An improved particle swarm optimization algorithm HPSO, which is based on human learning behavior, can effectively increase the population diversity, convergence speed and accuracy by changing particles' flight schema. In this paper, the HPSO algorithm is theoretically analyzed from the perspective of discrete-time linear system, the relevant criteria for guaranteeing the convergence of the HPSO are obtained. It provides a theoretical basis for the parameter setting in the algorithm, and reduce the dependence of the HPSO algorithm on parameters by enhancing randomness of the parameters. The experimental results based on the derived criteria confirm the performance.