Nowadays, scientific and efficient sports training is receiving increasing attention, especially personalized training monitoring. In this context, the development of wearable motion monitoring systems using advanced Convolutional Neural Network (CNN) technology can provide real-time and accurate training feedback for athletes, thereby optimizing training effectiveness. This study designed and implemented a wearable motion training monitoring system based on CNN, using multi-sensor fusion technology to collect physiological and motion data of users (such as heart rate, velocity, acceleration, etc.), and input these data into an optimized CNN for real-time analysis. In terms of CNN model design, this paper adopts a multi-layer feature extraction network, utilizing deep learning technology to accurately identify motion patterns and evaluate motion effects. The system also includes data preprocessing and feature extraction modules to improve data accuracy and model generalization ability. Regarding heart rate monitoring errors, the range of error values is between − 2 and 2. The application of this system not only improves the scientific and personalized level of training, but also provides intuitive data support for coaches and athletes, helping to develop more reasonable training plans and health management strategies.