52 0

Pose Estimation and Detection for Event Recognition using Sense-Aware Features and Adaboost Classifier

Title
Pose Estimation and Detection for Event Recognition using Sense-Aware Features and Adaboost Classifier
Author
김기범
Issue Date
2021-01
Publisher
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
Citation
2021 International Bhurban Conference on Applied Sciences and Technologies (IBCAST), Page. 500-505
Abstract
To examine events identification and recognition in sequential images, approaches used several parameters such as size, location or position of the human body parts along with its surrounding effects. In this paper, we estimated several body key points to monitor and track its appearance in complex events. Such detection of key body parts need several features descriptor such as optical flow, moving body parts and 0-180° intensity characteristics for event identification. While, extraction architecture model used human body key points through the R-Transform factor having intensity values. After intensity values extraction, the estimation of moveable body parts are applied for change detection system. In addition, optical flow features are also extracted. The extracted intensity and moveable body features are merged with optical flow vectors. These extracted features are injected into pre-classifier as Particle Swarm optimization and recognizer engine as Adaboost. The experimental results on a challenging videos datasets such as UCF 101 and YouTube show significant accuracy 75.33% and 76.66% of event recognition. We achieved higher performance of body-parts detection and event recognition as compare to state-of-art approaches.
URI
https://ieeexplore.ieee.org/document/9393293?arnumber=9393293&SID=EBSCO:edseeehttps://repository.hanyang.ac.kr/handle/20.500.11754/167047
ISBN
978-1-6654-0516-4
ISSN
2151-1411
DOI
10.1109/IBCAST51254.2021.9393293
Appears in Collections:
ETC[S] > 연구정보
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML


qrcode

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

BROWSE