A Survey of Human Activity Recognition in Smart Homes Based on IoT Sensors Algorithms: Taxonomies, Challenges, and Opportunities with Deep Learning | JoVE Visualize

A Survey of Human Activity Recognition in Smart Homes Based on IoT Sensors Algorithms: Taxonomies, Challenges, and Opportunities with Deep Learning

Damien Bouchabou^1,2, Sao Mai Nguyen¹, Christophe Lohr¹

¹IMT Atlantique Engineer School, 29238 Brest, France.

Sensors (Basel, Switzerland)|September 28, 2021

View abstract on PubMed

Summary

Smart homes use Internet of Things (IoT) sensors for human activity recognition to aid residents. This paper reviews current methods, challenges, and future research directions in this evolving field.

Area of Science:

Computer Science
Artificial Intelligence
Ubiquitous Computing

Background:

Smart home technology, driven by Internet of Things (IoT) advancements and reduced sensor costs, aims to enhance resident quality of life, particularly for the elderly and dependent.
Effective smart home services rely on accurately understanding resident daily activities through sophisticated recognition techniques.
Human activity recognition (HAR) in smart homes is a rapidly developing field facing continuous innovation and emerging challenges.

Purpose of the Study:

To provide a comprehensive overview of recent algorithms, research works, challenges, and taxonomy in human activity recognition using ambient sensors within smart home environments.
To identify and discuss specific problems, missing contributions, and future research needs in the nascent field of smart home HAR.
To propose directions, research opportunities, and potential solutions to accelerate progress in smart home activity recognition.

Main Methods:

Review and synthesis of existing literature on human activity recognition algorithms and techniques in smart home settings.
Analysis of current challenges and limitations in the field, including sensor data interpretation and activity classification.
Taxonomic classification of HAR approaches based on sensor types, algorithms, and application contexts.

Main Results:

Identification of key algorithms and methodologies currently employed for human activity recognition in smart homes.
Detailed examination of prevalent challenges, such as data sparsity, sensor heterogeneity, and real-time processing demands.
A structured taxonomy categorizing existing research and highlighting research gaps.

Conclusions:

Human activity recognition in smart homes is crucial for developing assistive services but faces significant challenges.
Further research is needed to address data integration, privacy concerns, and the development of more robust and adaptable recognition models.
The field presents numerous opportunities for innovation in algorithms, sensor fusion, and the creation of truly intelligent and supportive smart home environments.