Title: A review on learning-based algorithms for human activity recognition

Authors: Richa Mishra

Addresses: Department of Electronics and Communication, University of Allahabad, Prayagraj, India

Abstract: Human activity recognition (HAR) is an essential system that helps to identify, track and recognise human activities with the help of learning-based approaches in diverse areas of HCI, surveillance, health monitoring system, etc. The performance of the system is limited due to several factors such as placement of sensors, overfitting and underfitting of machine learning models, privacy issues, tracking GPS-based activities, cluttered background, etc. However, several comprehensive review papers have been published focusing on the state-of-the-art approaches using learning-based algorithms either in sensor-based HAR, vision-based HAR or both. This paper focuses on the research work to improve performance of the system using learning-based techniques focusing on video-based, sensor-based and Wi-Fi-based HAR. In addition, we have discussed the current state-of-the-art approaches, their merits and limitations, focused on the public datasets used for the performance evaluation, challenges and future directions for HAR.

Keywords: human activity recognition; vision-based; sensor-based; Wi-Fi-based; deep learning; machine learning.

DOI: 10.1504/IJDATS.2023.136681

International Journal of Data Analysis Techniques and Strategies, 2023 Vol.15 No.4, pp.339 - 355

Received: 12 Sep 2022
Accepted: 12 Nov 2023
Published online: 15 Feb 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article