Along with the rapid development of informationization in the medical industry and the increasing awareness of people’s health care, wearable monitoring technology has ushered in a golden period of growth, and in this context, a wearable athlete’s health monitoring device based on a radial basis function-probabilistic hybrid neural network (RBFNN) is designed. In this paper, the RBFNN is first used to identify the parameters and make corrections according to the changes of the controlled system. Then this neural network is optimized by the EM algorithm, and the EM-RBFNN algorithm that can optimize the smart wearable device is proposed. Through experimental comparison, although both RBFNN and EMRBFNN can match each sample to the number corresponding to the human health condition, EM-RBFNN has a higher accuracy in monitoring human health. The accuracy rate increased to 98%. Finally, through the rate analysis, blood oxygen and heart rate number reliability test, and motion misclassification rate test of the smart wearable device equipped with EM-RBFNN system, the smart wearable device installed with EM-RBFNN system is basically the same as the human body’s various data collected by the standard medical monitoring device. At a sampling rate of 50Hz, the real-time data acquisition rate increased by approximately 126%. The data on blood oxygen and heart rate have small errors. Smooth exercise and strenuous exercise have an error rate of between 10-20%, which is within the normal error range. Therefore, the smart wearable device based on EM-RBFNN can comprehensively monitor the health status of athletes.