ABSTRACT
Human activity recognition is one of the popular fields of research in the last two decades. This activity recognition topic is inspired by many useful real-world applications, such as simulation, visual surveillance and understanding human behavior. Till today, there are so many research studies going on in this field. This research has become one of the important and necessary one in the area of computer vision. Any small movement or change around us is directly or indirectly related to an activity performed by a person, and also, most of the activities are captured in the form of images and videos. So, there might be some situations where we have to identify the activity from these images and videos. Understanding human behavior from images and videos gives useful information for many computer vision problems and has many applications like scene recognition and human-computer interaction. There are many actions that a person can perform in his/her day-to-day life. Those actions can be either atomic or non-atomic. The actions like sitting, standing, laying, running and walking are some examples of atomic actions. Atomic actions are also called Action Primitives. Non-atomic actions are combinations of atomic actions. There are many systems that can recognize the activity using the smartphones. But, recognizing activity from videos or images is a challenging task due to variations in lighting, viewpoint, scale etc. In this chapter, we will develop a system that recognizes some of the atomic actions from images.
