An IoT and machine learning enhanced framework for real-time digital human modeling and motion simulation

Abstract

The Digital Human Model (DHM) is a computational technique employed for simulating human motion and behavior, finding applications across sectors such as manufacturing, healthcare, and education. Despite its potential, current DHM simulations grapple with challenges like high computational costs, significant latency, low precision, and limited realism. Addressing these challenges, this paper introduces a novel DHM simulation approach that synergizes the Kinect motion capture system, Internet of Things (IoT) devices, and advanced machine learning techniques. Distinctively, our method captures real-time human motion data through IoT devices and the Kinect system, followed by data preprocessing and feature extraction. In contrast to traditional methodologies, we harness deep learning to establish time-series models, integrating convolutional neural networks and dilated convolutional kernels to amplify processing capabilities and cater to multi-scale data. Furthermore, we employ virtual reality technology to craft a digital human structural model, streamlining motion simulation and optimization. Experimental evaluations underscore that our approach consistently outperforms conventional methods across all metrics, demonstrating robust real-time simulation capabilities. In essence, this paper pioneers an innovative technological paradigm for DHM simulation, heralding new avenues for research and applications in related domains.