Efficient Human Pose Recognition Using Improved OpenPose Algorithm on Embedded Systems with RK3399

In order to address the limitations of traditional human pose recognition methods due to algorithms and hardware devices, which cannot meet the requirements of real-time and accuracy, a human pose recognition system based on convolutional neural networks was designed and implemented using the RK3399 chip to improve recognition accuracy and real-time performance. Adopting a deep convolutional neural network structure, including multiple convolutional layers, pooling layers, and fully connected layers, for feature extraction and classification. Through a large dataset and training, the model is able to accurately recognize human postures at different angles, lighting conditions, and backgrounds. The experimental results show that the method has high accuracy and robustness in human posture recognition, and can be applied to human-computer interaction, motion capture and other fields. The convolutional neural network human posture recognition method based on RK3399 proposed in this paper provides a new solution for the application of embedded systems in the field of human posture recognition.Although this leads to a decrease in recognition accuracy, improving OpenPose greatly reduces the amount of computation, greatly reduces the burden on the graphics card, and also reduces the waiting time for computation and improves the timeliness of detection.