CSEar: Metalearning for Head Gesture Recognition Using Earphones in Internet of Healthcare Things

With the popularity of personal computing devices, people often keep long-term head immobility in front of screens, resulting in the emergence of "phubbers" and "office workers." The early warning solutions in the Internet of Healthcare Things (IoHT) have brought hope to protect...

Full description

Saved in:
Bibliographic Details
Published in:IEEE internet of things journal 2022-11, Vol.9 (22), p.23176-23187
Main Authors: Bi, Hongliang, Liu, Jiajia
Format: Article
Language:English
Subjects:
Accelerometers
Adaptation models
Algorithms
Data models
Earphones
Gesture recognition
head gesture
Headphones
Health care
Internet
Internet of Healthcare Things (IoHT)
Internet of Things
Legged locomotion
metalearning
Neck
Traffic accidents
Wireless communication
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:With the popularity of personal computing devices, people often keep long-term head immobility in front of screens, resulting in the emergence of "phubbers" and "office workers." The early warning solutions in the Internet of Healthcare Things (IoHT) have brought hope to protect users' health and safety. However, most existing works cannot recognize the different head gestures during walking, which is also a common cause of text neck and traffic accidents. In addition, they also need a large amount of data to update the model to adapt to the new environment, which reduces the practicality of the model. To solve these problems, we propose a system, CSEar, based on built-in accelerometers of off-the-shelf wireless earphones, which can recognize 12 kinds of head gestures both in resting and walking states. First, an innovative algorithm is designed to detect head gesture signals, especially for the signals mixed with gait. Then, we propose the MetaSensing, a head gesture recognition model that can improve the recognition ability with few samples compared with the existing metalearning algorithms. Finally, the experimental results prove the effectiveness and robustness of the CSEar.
ISSN:2327-4662
2327-4662
DOI:10.1109/JIOT.2022.3188331