An Efficient Model for Identifying Human Behaviour using Machine Learning

The extensive domain of Human Activity Recognition (HAR) is dedicated to the identification of specific human actions or movements through the use of sensor data. Furthermore, it is capable of detecting and distinguishing human activity, as well as disseminating information about the detected motion or activity. Skeleton coordinates are employed by the HAR program to identify human body activity or motion. To be more precise, this application serves as an instrument for the analysis of human activity and is capable of automatically detecting the following activities: standing, sitting, lying down, walking, stairs, and walking. This paper introduces a hybrid model that employs deep learning to detect human activity. The model includes a diverse array of layer counts and the six most common activities. Furthermore, the proposed model will recognize behaviours such as walking, standing, sitting, lying down, falling, walking, and walking on staircases. The individual's activities will be consistently monitored by the proposed model, which will also issue appropriate notifications. However, this model is capable of detecting human activity in real time using even the most basic computer, in contrast to other models that require extremely potent CPU specifications to function in real time and rely on sensors to detect activity. Therefore, this paper introduces a novel hybrid deep learning model for Human Activity Recognition (HAR) that effectively detects six common activities using skeletal coordinates on fundamental computing systems without the need for sensors. It operates across a variety of camera angles without the need for code adjustments, as it seamlessly integrates with CCTV systems. The model is deployable online, guaranteeing accessibility through web browsers, and its accuracy can be improved by utilising real-world datasets such as CCTV footage. This method expands the applicability of HAR to a wider range of applications, including healthcare, smart homes, surveillance, sports, and more, with minimal computational requirements.