Device-Free Human Activity Recognition: A Systematic Literature Review

Abstract

Human activity recognition (HAR) has become a topic of interest in recent years. While device-based, object-tagged, and camera-based approaches to HAR have many advantages, device-free HAR offers new contributions to the field. Unlike device-based or object-tagged methods, it does not require users to carry sensory devices, and unlike camera-based methods, it respects privacy. Despite the significant number of original research studies and surveys on device-free HAR published in recent years, many scientific questions remain open. In this study, a systematic literature review on device-free HAR was conducted by exploring ACM, IEEExplore, ScienceDirect, Scopus, and WebOfScience. This mixed-method study assesses the quality of the reviewed papers and analyzes suggested HAR methods in both a scientometric and technical manner. The scientometric analysis investigates the trends of scientific publications in this field from the beginning of 2017 to the end of 2023 and reviews the types and distribution of publications among countries, universities, and media. The technical analysis categorizes methods based on device-free sensing modalities, the type, and granularity of recognized activities of proposed methods. It also discusses the common challenges and limitations of current device-free HAR approaches. Additionally, existing methods are compared based on their support for non-line-of-sight, multisubject, user-independent, and environment-independent recognition of human activities. This work provides foundational knowledge on each step of device-free HAR: data acquisition, preprocessing, classification, and evaluation, and identifies gaps and open questions in existing research.