![]() | Only 14 pages are availabe for public view |
Abstract IoT systems generate vast amounts of sensor data from diverse devices, requiring computationally intensive processing. Cloud servers are commonly employed for real-time data processing, but the distance between servers and data sources can introduce transmission delays. To mitigate latency and reduce costs, this research focuses on deploying machine learning (ML) models at the edge, closer to the data source, to minimize data traffic to the cloud while maintaining accuracy. The primary objective of this study is to enhance the performance of distributed intelligence of IoT networks in human activity recognition (HAR) systems. The research employs distributed computing across interconnected devices and sensors in IoT networks, leveraging different ML models to analyze and recognize human activities. Instead of relying on a centralized system, the intelligence is distributed among devices and sensors, enabling realtime analysis, improved scalability, and enhanced decision-making capabilities. HAR systems are crucial in healthcare, sports, and rehabilitation, requiring continuous monitoring and rapid responses. ML at the edge is essential for quick and accurate activity recognition. While cloud computing offers high accuracy, the traffic and delay between the cloud and end systems can introduce challenges. This thesis utilizes smartphones and wearable sensors fo |