The complex system of human body can be observed by observing the change of gait rhythm.The dynamic characteristics of the time series of stride interval during walking can effectively reflect the state change of human system, which can be used for abnormal gait detection and related disease identification.Time series phase space reconstruction of human gait sensor information is an effective modeling method to characterize nonlinear dynamics of the system. Geometric modeling and statistical analysis of phase space are typical analysis methods for abnormal gait recognition, which are widely used in clinical research such as neurodegenerative disease detection. In this paper, a topological nonlinear dynamic modeling method is proposed to identify abnormal gait in neurodegenerative diseases from the perspective of spatial topological analysis. Firstly, a phase-space reconstruction method based on time-delay embedding is used to transform the wave time series of gait into an abstract phase-space state point cloud. Secondly, the continuous homology tool based on computational topology is used to extract the topology description information of the space where the state point cloud resides. Thirdly, the topological nonlinear dynamic characteristics of time series are constructed by using the continuous situation graph based on topological description. Finally, a machine learning recognition model of abnormal gait was constructed by integrating the topological nonlinear dynamic characteristics of the time series of stride interval, standing interval and swing interval of left and right feet in the gait cycle as the input of classifier. The results showed that the area under receive-operator-curve was 0.875 0 (0.914 6), 0.940 6 (0.962 3) and 0.958 3 (0.961 4) in the 5-minute continuous walking data (50-step sliding window data) of abnormal gait in patients with neurodegenerative diseases of muscular sclerosis, Huntington’s disease and Parkinson’s disease, respectively. Therefore, topological nonlinear dynamic modeling analysis is an effective gait detection method for neurodegenerative diseases, and provides a new idea for neurodegenerative disease detection and wearable data analysis based on gait analysis.