Artificial Intelligence-Assisted Pattern Discovery for Narrow Domains of Human Activity Using a Heuristic-Optimised Approach

This work outlined the process of creating automatic recognition systems for analysing human behaviour. Here, heterogeneous sensors are implanted in the human body to record data in real-time, enabling monitoring of a wide range of physiological signals in response to the subject's behaviour. In addition, this study aimed to examine the ages and daily routines of people with physical disabilities who rely on ambient assisted living technology. This study analyses a state-of-the-art Human Activity detection (HAR) approach by analysing signals collected from real-world activity detection applications. For activity recognition, this research employs a deep belief network trained in an unsupervised layered manner on statistical features extracted from multi-sensor a smartphone data. Data from the human sensors are gathered via high-tech mobile devices and then analysed by a robust deep-learning network. The experimental study shows that several standard approaches have been used to analyse the collected data based on the HAR pattern at lab size. The suggested technique is evaluated using the HAR datasets, and the findings demonstrate that it has a higher accuracy (98.67%) than the standard system.