Human activity recognition and authentication system

Abstract

Automated surveillance motion detection and video data analysis are now crucial jobs for many industries. Understanding human behavior from security video data is crucial, especially in places like banks, hospitals, superstores, and other restricted areas. The two most discussed subjects in the field of computer visions are face detection to identify people and human activity recognition. Over the past 20 years, numerous study projects have been conducted. I’ll discuss the Human activity Recognition and Authentication (HARAuth) System initiative in this essay. In this project, I’ll suggest an algorithm to identify human activity while also authenticating the individual to determine whether that person is authorized to perform that activity. In this work, I presented a method for classifying and recognizing particular activities based on the pose skeleton of a human. Pose estimation and classification are the first two steps in this procedure. This project uses the OpenPose library for its pose estimation tasks. Additionally, MLPClassifier from the Sklearn library is used to complete the activity classification job. I cropped each person’s rectangular area during the pose classification process based on the pose’s position in the frame-by-frame video image. Each person’s rectangular area is subjected to face recognition in order to verify their identity for the identified action.