Under the advocacy of “Healthy China”, people pay more and more attention to sports. However, the traditional motion monitoring system is still limited in the aspect of posture information monitoring, especially in the aspects of limited capture range, data loss, and inaccurate motion recognition. Based on several Kinect devices, combined with Taijiquan and other sports, this paper explores a new way of sports posture monitoring. The research covers Kinect technology, bone tracking, data fusion, movement recognition and other fields. The system first collected human bone information under multiple Kinect, and extracted key bone point data through calibration and fusion technology, which was used as the basis for subsequent action recognition. The real-time classification and recognition of motion is successfully realized by using the extracted real-time bone coordinates and Angle features and the dynamic time warping algorithm. Aiming at the shortcomings of existing monitoring systems, this paper proposes a real-time recognition system of sports images based on multi-sensor fusion. Through the joint deployment of angular velocity and inertial sensor, combined with the wavelet technology to process the data, the system has achieved a callback rate of more than 90% and a response within 0.51 s, which brings broad application prospects for sports posture information monitoring. The aim of this study is to provide effective monitoring and guidance for physical activity, reduce risks, optimize outcomes, and thus provide reliable support for healthy physical activity participants.