AIARS: Towards Automated Infant Activity Recognition System using Machine Learning and Multi-sensor Fusion Data

Date

2025-01-07

Contributor

Advisor

Department

Instructor

Depositor

Speaker

Researcher

Consultant

Interviewer

Narrator

Transcriber

Annotator

Journal Title

Journal ISSN

Volume Title

Publisher

Volume

Number/Issue

Starting Page

3254

Ending Page

Alternative Title

Abstract

Infant Activity Recognition (IAR) plays a pivotal role in modern healthcare and developmental studies,offering crucial insights into early childhood behavior and motor development. Nevertheless, identifying various categories of infant activities such as playing,crawling, and feeding, poses challenges due to the dynamic nature of a child’s age and the continuous involvement of parents or caregivers. Recognizing infant activities aids in the timely detection of motor disorders and promotes healthy movement behavior. In this study,we propose an automated Infant Activity Recognition System (AIARS) utilizing a Machine Learning (ML)approach that leverages data from wearable sensors.Initially, we compiled an Infant Activity Database(IAD) by collecting data from infants across 16 distinct activities in controlled laboratory environments. Due to the limited sample size for each activity, we aggregated these 16 activities into 3 and then 2 broader categories.Employing ML techniques, including Random Forest(RF), Support Vector Machines (SVM), and Extreme Gradient Boosting (XGB), we developed an AIARS to classify various infant activities. Our study achieved impressive results, with an accuracy of 93.85% , an Area Under the Curve (AUC) of 99%, and an F1 score of 95.2%. We also addressed the challenges faced while applying ML methods in developing AIARS and provided recommendations to mitigate these challenges.These findings underscore the efficacy of our approach and represent a significant milestone in the domain of IAR.

Description

Keywords

Body Sensor Networks for Personalized Medicine, accelerometer, infant activity classification, infant activity recognition, machine learning, physical activity

Citation

Extent

10

Format

Geographic Location

Time Period

Related To

Proceedings of the 58th Hawaii International Conference on System Sciences

Related To (URI)

Table of Contents

Rights

Attribution-NonCommercial-NoDerivatives 4.0 International

Rights Holder

Local Contexts

Email libraryada-l@lists.hawaii.edu if you need this content in ADA-compliant format.