A Study on Handcrafted Features Based Models in Human Actions Recognitions

An Investigation into the Efficiency of Handcrafted Features for Human Actions Recognition in Video Sequences

Abstract

The vision-based comprehension in video sequences entices several real-life applications suchas gaming, robots, patients monitoring, content-based retrieval, video surveillance, and security. One of theultimate ambitions of artificial intelligence society is to produce an autonomous system that can beidentified and interpret human behavior and activities in video sequences properly. Over the decade,numerous efforts are made to detect the human activity in films but nevertheless, it is a tough work owingto intra-class action similarities, occlusions, view variations and ambient factors. These methods aredivided into handwritten features based descriptors and automatically learned feature based on deeparchitectures. The suggested action recognition framework is separated into handmade and deeplearning-based architectures which are then employed throughout this study by incorporating the novelalgorithms for activity detection.