Continuous Person Recognition on Balloon Robot Using TSK Fuzzy Classifier with Face and Body Regions for Social Distance Purposes

The coexistence of humans and robots is the current trend. One of the key factors to discuss is the social distance between humans and robots. Instead of humans maintaining physical distance from robots, robots should do so, especially pet robots to maintain varying social distances between stranger...

Full description

Saved in:
Bibliographic Details
Main Authors: Siow, Chyan Zheng, Wang, WeiHao, Obo, Takenori, Kubota, Naoyuki
Format: Conference Proceeding
Language:English
Subjects:
Balloon Robot
Body regions
Cameras
Face recognition
Feature extraction
Fuzzy Model
Growing Neural Gas
Human factors
Legged locomotion
Person Recognition
Robot vision systems
Social Distance
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The coexistence of humans and robots is the current trend. One of the key factors to discuss is the social distance between humans and robots. Instead of humans maintaining physical distance from robots, robots should do so, especially pet robots to maintain varying social distances between strangers or owners. In this paper, we propose a balloon robot (pet robot) that can recognize person walking towards it. However, most research on person recognition has been done under the condition that faces are detected. In addition, in order to use the body for recognition, they require the user to maintain a certain distance from the robot camera. Therefore, when a person comes very close to the robot, the robot will not be able to recognize the person. In this paper, we first divide person recognition into face recognition and body recognition, and then use the TSK fuzzy model to fuse these recognition results. If the recognition is accepted, we push the extracted features to the GNG cache system for subsequent recognition. Additionally, we introduce the use of GNG cache system placed on an image grid to extract body region when the person is very close to the camera. We installed a camera on the top of a mobile balloon robot and used it to collect video data from 10 college students. Through experiments, using the GNG model as a caching system can improve the recognition results. Image grids are learned through two GNG models and are able to extract body regions from the frame when a person stands very close to the robot. Ultimately, the proposed method was able to successfully recognize person with 90 % accuracy when using the TSK fuzzy model.
ISSN:1558-4739
DOI:10.1109/FUZZ-IEEE60900.2024.10612063