Usage of IoT, High-Performance Computing, and Machine/Deep Learning in Human Activity Recognition Systems
Elangovan Ramanujam, P Meena Kumari, K Bharathi · Auerbach Publications eBooks · 2025
The human activity recognition (HAR) system detects simple and complex human activities in the smart home by processing spatial and temporal information acquired by visual and non-visual sensory data. Visual cameras such as infrared (IR) and depth (RGB-D) are mainly used to collect visual data. Wearable sensors/devices, smartphones, and ambient sensors are used to collect non-visual sensory data. Recently, ambient sensors, especially Internet of Things (IoT) devices/sensors, in the elderly living environment, such as a wall, chair, or table, have created a milestone in activity recognition. These sensors are more advantageous than wearable ones because they can indirectly indicate elderly activities for their caretakers or family members in an emergency without causing panic in the elders. Various researchers have shown interest in IoT-centric HAR systems in recent years. They have shown promising results due to the evolution of deep learning models and hybrid machine learning techniques. However, most of the system concentrates on single-resident activity, which is simple and more accessible. In real-world smart home automation, a single resident is not always the case; thus a multi-resident activity recognition system is required. This multi-resident activity recognition system produces more sensor data, requiring high-performance computing (HPC) systems to handle the generated data. Very few research works have introduced the concept of HPC for HAR systems. Thus, this chapter overviews the research works proposed for the HAR system using HPC to analyze the data generated by IoT devices/sensors through deep/machine learning with scope for opportunities.