Information plays a crucial role in the reform of public physical education courses in colleges and universities. In this paper, we construct a sports health monitoring system for teaching physical education in colleges and universities. Firstly, we derive the heart rate and transmittance oxygen saturation monitoring algorithm using Lambert’s law and the photoelectric volume method, and then design the anti-interference heart rate and oxygen extraction algorithm using a time-varying autoregressive model. To use the time-varying autoregressive model based on multiwavelet basis function expansion, you have to pick the multiwavelet basis functions and guess what the model parameters are. Then, the action recognition module is designed, the static calibration method is used to correct the zero bias of the collected data, the sliding mean filtering method is used to eliminate the random noise, and the data segmentation is performed on the long sequence data. After the system is constructed, its accuracy is checked in four states, and finally, its effect on students’ exercise health is analyzed. It was found that under the four conditions of sitting, standing, walking, and jogging, and comparing with the results of medical pulse oximetry, the maximum error of the heart rate and blood oxygen results monitored by the system in this paper is no more than 3, which is within the acceptable range. The coefficients between the use of the exercise health monitoring system and the enthusiasm for exercise, commitment to exercise, and adherence to exercise health were 0.571, 0.584, and 0.625, respectively, with a significant correlation (p<0.01) and a moderate correlation between the variables. Exercise enthusiasm and commitment can both serve as simple mediators in the relationship between the use of the exercise health monitoring system and exercise health adherence, and they can also act as a chain mediator in the relationship between the use of the system and physical activity adherence.