Automated Patient Activity Identification in Cyber-physical Systems Using A Unique Deep Learning Approach and Multi-objective Optimization

Abstract

Aims and Background: For video understanding and analysis, human activity recognition (HAR) has emerged as a challenging field to investigate and implement. Patients can be monitored in real-time by a group of healthy individuals, and abnormal behaviors can be used to identify them later. Patients who do not engage in customary physical activities are more likely to suffer from stress, cardiovascular disease, diabetes, and musculoskeletal disorders. Thus, it is critical to collect, evaluate, and analyze data to determine their activities. Objectives and Methodology: Deep learning-based convolutional neural networks (CNNs) can be used to solve the problem of patient activities in the healthcare system by identifying the most efficient healthcare providers. Healthcare relies heavily on the integration of medical devices into cyberphysical systems (CPS). Hospitals are progressively employing these technologies to maintain a high standard of patient care. The CNN-CPS technique can be used by a healthcare organization to examine a patient's medical history, symptoms, and tests to provide personalized care. A network of medical devices must be integrated into healthcare. Hospitals are increasingly using these technologies to ensure that patients get the best possible care at all times. Healthcare automation can dramatically improve quality and consistency by reducing human errors and fatigue. The multiobjective optimization is achieved considering various factors like the time required to find emergency case identification, new disease prediction, and accuracy of data protection. Results: Consequently, patients are assured of receiving a consistent, attentive service at every visit. Data and orders can be stored and entered more easily via automation, market research suggests. The outcome of this article is obtained based on a comparison of various approaches in health monitoring systems, such as collection of patient data is 82.3%, new disease prediction is 80.14%, emergency case identification is 78.25%, data protection is 81.35%, immune to impersonation attack reduction is 78.36% and overall accuracy of data protection and transmission performance is 86.24% is achieved. Conclusion: Compared with existing methods DM-CC and HE-WSN for health monitoring and patient’s treatment process, the proposed method CNN-CPS is better in maintaining the data and proper information passed to the medical care is 92.56%.