Deep learning is a capacity to offer a comprehensive approach to automatic feature extraction and predictive modeling for analysis and decision-making has led to the field of deep learning becoming more and more popular in the Internet of Things. This work presents a Deep Learning Framework-based Internet of Things dance movement recognition model. To recognise dancing movements from the collected data by an Internet of Things device, the framework uses a convolution neural network (CNN) with a data-centric architecture. Three accelerometers record 3D motion data for the Internet of Things gadget. The CNN architecture is then used for feature extraction, producing a flattened matrix that depicts movement. A Multi-Layer Perception (MLP) is then employed to categorise the movements. A standardised dataset of sixteen dance steps with three speed settings is used to empirically assess the suggested solution. The outcomes demonstrate that, in terms of classification accuracy, evaluation time, and accuracy, our model performs better than cutting-edge methods. With 90.74% accuracy, 86% precision, 84% recall, and 89% F1-Score, the suggested model performed well. The suggested model can be the foundation for an accurate and user-friendly system that tracks patients’ dancing movements.