Human Handover Classification using a Deep Learning Model

Detection and classification of handover actions is one of the new challenging tasks. Robots must be able to detect and differentiate between different ways of handover in order to make the correct move. In this paper, a combination of deep learning architectures, an LSTM network, and feature select...

Full description

Saved in:

Bibliographic Details
Published in:	2021 31st International Conference on Computer Theory and Applications (ICCTA) pp. 187 - 192
Main Authors:	Monir, Islam A, El-Bendary, Nashwa, Fakhr, Mohamed W
Format:	Conference Proceeding
Language:	English
Published:	IEEE 11-12-2021
Subjects:	Action Recognition Computational modeling Feature extraction Handover Human Handover Inertial sensors Multimodality Receivers Robot sensing systems Robot-Human Interaction Robotic Grasping Robotic Handing Tracking
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	Detection and classification of handover actions is one of the new challenging tasks. Robots must be able to detect and differentiate between different ways of handover in order to make the correct move. In this paper, a combination of deep learning architectures, an LSTM network, and feature selection techniques are employed to classify human handovers from both the giver and receiver perspectives. A publicly available dataset consisting of motion tracking sensor measurements, Kinect readings for 15 joint locations, 6-axis inertial sensor readings and video recordings was employed in the evaluation. The results came to be promising as 91% accuracy and precision with 88% recall are obtained using the proposed model.
ISSN:	2770-6575
DOI:	10.1109/ICCTA54562.2021.9916602