In recent years, with the rapid development of Internet of things and other technologies, the digitalization, networking, and intelligence of sports have become the current research focus. In this paper, the fitness management system based on the Internet of things is studied. By analyzing the system function and performance requirements, the design of fitness client (small tablet) of networked fitness management system is based on Internet of things. Receiving the fitness data uploaded by the fitness device through Bluetooth, the fitness data can be processed and displayed in real time with graphics. After the exercise, the fitness data can be uploaded to the central computer through Wi-Fi wireless. Taking barbell as an example, by analyzing the movement characteristics of barbell, Bluetooth MPU6050 module is used for data acquisition; the data collected includes angle, number, etc.; the relevant functions of barb-dumbbell movement are analyzed and designed; and the Bluetooth communication module and Wi-Fi communication module in the small tablet software system are designed and implemented. The relevant experiments were carried out based on the developed software and hardware platform. Recognition experiments on 7 classes of actions show that the proposed deep neural network learns well on small datasets, achieving an action recognition accuracy of 97.61%, and the SVM also achieves a recognition accuracy of more than 96%. In the 50 action cycle calculation experiments, the number statistics algorithm reached 100% calculation accuracy, and the action cycle calculation results are also close to the real value, proving the effectiveness of the periodic calculation method.