In order to solve the problem of poor realtime performance and low recognition accuracy in pedestrian multidirectional walking pattern recognition, a multidirectional walking recognition algorithm based on acceleration timedomainslope feature is proposed in this paper. By calculating and analyzing the magnitude of the acceleration slope of the pedestrian motion direction at the zaxis acceleration peak or trough, the walking pattern of pedestrians moving forward, backward, left shift, and right shift in multiple directions can be accurately identified. The results show that the algorithm can recognize pedestrian walking patterns in real time, and the average recognition accuracy is above 96%.