<p>The trajectory recognition of personnel in coal mines refers to accurately linking the target trajectory to the corresponding personnel. Due to the complex environment in coal mines, the trajectories of personnel in coal mines have problems such as sparse and uneven trajectories. Most of the existing methods only focus on the relationships within the trajectory sequence itself and do not fully consider the impact of the time interval between trajectory points on trajectory encoding, resulting in a relatively low accuracy of trajectory recognition. This paper proposes a method for identifying personnel in coal mines based on temporal perception self-attention trajectory encoding (Temporal Perception Self-attention trajectory Encoding for Coal mine underground personnel Identification, TPSECI). This method designs multiple trajectory feature embedding modules to extract the embedding representations of trajectories in terms of space, time, and workplace. Among them, it contains Temporal Periodicity Feature Embedding Module, which solves the problem that most existing research methods ignore the influence of time interval between trajectory points on encoding. At the same time, a temporal decay self-attention trajectory encoding module is designed to perform deep encoding representation of the trajectories. Experimental results on a trajectory dataset of a certain coal mine show that the performance of TPSECI is, on average, 3.37%, 2.4%, and 2.56% higher than that of the benchmark model AttnTUL in terms of ACC@K1, ACC@K5, and Macro-F1.</p>