Real-Time 3D Human Tracking and Pose Construction Using Millimeter-Wave Radar System

Overview

Deviceless human recognition offers excellent potential for human-machine applications in healthcare and intelligent environments. Although current methods using WIFI Channel State Information (CSI) achieve promising results in controlled lab settings, real-world applications are still limited due to issues with model training and rapid environmental changes. Millimeter-Wave (mmWave) bands have gained interest as an indispensable tool for high precision localization, real-time domain independence, and high angular and spatial- temporal processing. This study explores the use of a portable mmWave device to construct human poses accurately in real-time. The system will identify the radar signal variation caused by the moving subject and analyze the 3D point cloud generated by each moving body part. A deep Graph Neural Network (GNN) takes the 3D point cloud structures in the spatial dimension, learns the spatial relations between each 3D point cloud, and predicts the human pose structure. Experimental results across subjects, environments, and locations demonstrated an average accuracy of 98.92% with a 3.3ms lag time and 2.23cm average error compared to a professional 3D Depth camera with advanced skeleton tracking SDK. These results validate using a device-free mmWave system for accurate human pose construction under complex domains and none-line-of-sight (NLoS) scenarios. An Exploratory and Control Study of the Endolysosomal Pathway in Alzheimer’s Disease Ambika Polavarapu Millburn High School, Millburn, NJ Alzheimer’s disease is the most prevalent neurodegenerative disease today with 44 million patients worldwide, according to the World Health Organization. A hallmark of the disease and one of the first pathologies seen is amyloid-beta plaque. The endolysosomal pathway is involved in the secretion of the plaque as it is responsible for protein trafficking in the cells, specifically from the surface to endosomes. This study focused on developing a methodology for targeting and labeling proteins in endosomal compartments through imaging of cell lines stained with antibodies, endosomal markers, and fluorophores to better understand the endolysosomal pathway. Results show through multiple colocalization analyses that protein targeting and labeling was successful. Mander’s Coefficients derived from the analyses indicated that the selected methodology correctly tracked the synthetic protein to the endosomes and labeled it efficiently. The results of this work can be used to manipulate the endolysosomal pathway in order to observe how endolysosomal dysfunction plays a role in the pathology of Alzheimer’s disease.

Competition history

• JSHS 2022 Category not listed

Resources

• National JSHS abstract book ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.