Advancements in Human Activity Recognition: A Comprehensive Survey of Sensor-Based and Vision-Based Approaches

The significance of human activity recognition (HAR) has increased in recent years because of its wide-ranging applications in areas such as healthcare, security and surveillance, entertainment, and intelligent settings. A significant challenge in computer vision is the automated and accurate recognition of human actions. This survey presents the last related work conducted over the 2015–2023 years in various areas of human activity recognition. It introduces the classification of HAR basic methodologies. In general, HAR approaches are categorized into two primary groups: sensor-based and vision-based HAR. This classification is based on the type of created data and the system environment itself. Based on our study, smart phones, smart watches, accelerometers, gyroscopes, and arm bands are all examples of sensor-based techniques. For the vision-based techniques, there are cameras, Microsoft Kinect, and thermal cameras. Also categorized the general datasets referenced in academic papers are based on the type of activity: individual actions, behaviors, interactions, and group activities. Subsequently, the preprocessing and feature engineering procedures are demonstrated. At long last, this review is able to offer some study concepts that investigate and analyze HAR.