With the aim of realizing accurate target tracking and improving the efficiency of target tracking for recreation-assisted robots, research on the target tracking technique for recreation-assisted robots was carried out by using a discrete Hidden Markov Model. First, multiple visual sensor devices built into the mobile robot are utilized to acquire target image data from multiple viewpoints, and the multimodal data acquisition results are obtained. Second, a particle filter is constructed to realize the detection of the tracking target by the mobile robot. A discrete Hidden Markov Model is established to infer the target state. On this basis, the target trajectory of the recreation-assisted robot is generated and tracked in real time. The experimental results show that after the application of this method, the target tracking error of the recreation-assisted robot is smaller, the average update delay of tracking is shorter, and it has better stability throughout the entire tracking process, enabling it to track the target continuously and stably.