Intelligent human activity recognition for healthcare digital twin

Human activity recognition and healthcare monitoring are becoming increasingly popular as cost-effective and innovative solutions to improve the standard of healthcare in the era of Industry 4.0. The concept of the Internet of Healthcare Things (IoHT) supports these solutions and builds a virtualized and software-controlled infrastructure. This new approach leads to the development of new concepts by digitalizing and connecting everything. Despite the significant advancements in IoHT, there are still challenges in processing vast amounts of data and handling resource-limited devices. In this regard, digital twin technology is an emerging tool to enhance IoHT services. With the help of digital twin, data processing at the edge devices can effectively overcome these challenges by reducing data transfer limitations and latency while improving prediction accuracy. In this paper, we present an intelligent human activity recognition framework in healthcare digital twin services. Our framework creates digital twins of wearable and portable devices/sensors in the physical network, collects real-time and historical data, and applies advanced analytics for feedback. The main contributions of this paper are: (i) We propose a novel four-layer digital twin architecture framework for human activity recognition. (ii) We discuss how the layered architecture and data processing at the edge devices enhance decision-making and classification accuracy. It is also aimed to design an environment where data with different characteristics, priorities, and transmission timings (i.e., regularly transmitted and critical) are comprised so that we can measure the same state through multiple sensors to improve system performance. (iii) We develop an Artificial Neural Network (ANN) based model and evaluate the proposed digital twin-assisted model using two different datasets. The results show the benefits of the proposed digital twin-assisted framework, providing feedback to individuals.