A Review on Human Action Recognition Approaches

Abstract

Action Recognition is a new world-specific branch implemented in various applications like surveillance, health monitoring, and computer vision. Human Action Recognition is a challenging task in the research area field. Most of the researchers suggested efficient machine learning HAR algorithms like SVM and KNN. The dataset used in state of the art HAR schemes is readily available on a social platform. The features learning and classification approaches are used for motion detection. Activities like walking, running, jumping, sleeping, falling, and interaction are recognized by the machine learning approach. In this paper, we are presenting various machine learning and hybrid algorithm for the HAR task. Different datasets tested by the researchers with their developed work. The accuracy achieved by all reviewed studies also mentioned with their advantages. Based on these studies, we work on the combination of deep learning and machine approaches for the HAR. The HAR approaches depend on the wearable sensor as well as the recording devices.