The separating behavior defines the division of a crowd from a single flow into two distributary flows due to the different pedestrians' destinations. Nevertheless, in the existing literature on pedestrian flow, there is a lack of simulation research on the separating crowd behavior in a T-shaped channel. By conducting a series of controlled experiments, we analyzed the moving trajectories and the spatial and temporal distribution characteristics of pedestrians in the separation process. Based on an analysis of the controlled experiments, we proposed an improved social force model that fully considers the characteristics of pedestrians' swapping locations, and refines the directions of pedestrians' expected speeds in three stages of the pedestrian separation process. During the simulation, we applied the improved model to explore the effects of the pedestrians' swapping locations on the macroscopic phenomena, microscopic individual behavior, and traffic efficiency within a T-shaped channel. The simulation results show that if pedestrians' swapping locations are concentrated in a certain area close to the entrance, the traffic efficiency in the T-shaped channel will be higher than that if the pedestrians' swapping locations are dispersed. Moreover, as the flow rate at the entrance increases, the swapping location becomes more concentrated closer to the entrance, the mean speed increases, and fewer conflicts occur between the pedestrians.