Advancements in Radar Point Cloud Processing for Macro Human Movements in Healthcare and Assisted Living Domains: A Review
Shahzad Ahmed, Sohaib Abdullah, Sung Ho Cho
Abstract
Point clouds (PCs) are ubiquitous data representation schemas in complex tasks related to semantic segmentation and scene understanding. Contrary to vision-based approaches, radars, being a privacy-preserving sensor, are lately getting huge attention in generating PCs for medical applications since such sensors can be embedded into hospitals and living spaces. This article summarizes the use of radar-generated PCs in the healthcare and assisted living domain. Comparative analysis of radar and other technologies is presented briefly, followed by a detailed note on commercial radars for PC generation. Radar PC data collection, preprocessing, feature extractions, and feature processing are reviewed, and a detailed summary of applications related to healthcare and assisted living is presented. Supporting signal processing and machine learning (ML) approaches are also reviewed. Specifically, the dedicated PC-oriented ML algorithms are discussed in detail. The discussed applications encompass human activity recognition (HAR), posture classification, gait recognition, and fall detection. Radar PC data are crucial for certain health monitoring and rehabilitation tasks, such as skeletal-joint and pose estimation; the range, Doppler, and angle information of the target independently may fall short in such applications. Finally, this article concludes with a comprehensive summary of current trends, key takeaways, and future directions. This article also outlines the future prospect of using generative ML for healthcare applications.