Vital Signs Based User Authentication Using mmWave Radar

Overview

The goal of this project is to use a single radar for contactless, continuous and unobtrusive biometric authentication (BA), which involves determining the identity of a user by comparing the user’s radio sensed data with existing biometric templates to determine their resemblance. The project includes radio data collection, random signal processing, data analysis and algorithm development. Specifically, the Radar is placed at a fixed location, pointing to the user to be authenticated. It sends out a low-power high frequency electromagnetic wave that is reflected by objects in its path. The hypothesis is that the reflected signal contains a user’s unique biometric information that can be extracted and matched to its biometric signature. In order to clean the signal that also picks up random body movements and environmental noise, the raw signal is processed by a number of procedures including empirical mode decomposition and digital signal filtering. Once the cleaned signal containing consecutive cardiac cycles is obtained, unique features of the cardiac motions are extracted to train a machine learning based algorithm for user classification, where each user belongs to a unique class. For a database consisting of a dozen users, experimental results show that the machine learning based algorithm is able to detect a user’s unique cardiac movement pattern, and the authentication accuracy is over 98%. Decentralized, Autonomous Drone Swarms for Real-Time Mapping Applications and Natural Disaster Relief Richard Lian duPont Manual High School, Louisville, KY Decentralized, autonomous drone swarms have great potential in mapping of an unexplored environment, search and rescue, and intelligence, surveillance, and reconnaissance (ISR) applications. Drone swarms can complete missions that are too dangerous or complex for humans and single drones to perform. Most recent works on autonomous drone swarms focus on centralized swarm coordination and navigation planning, while research on decentralized drone swarms is still a large gap due to the complexity of a multi-robot system and higher costs. Moreover, recent drone mapping applications widely use photogrammetry with an RGB camera due to its low-cost, lightweight, and color information, which fails in outdoor settings with inadequate illumination. Furthermore, photogrammetry methods have low topographical accuracy and depth measurement 51 precision, which results in unreliable digital terrain models. In our research, we propose a distributed system focused on coordinating the actions of an entire fleet by continuously updating the status of individual drones to neighboring drones. Additionally, low-cost LiDAR sensors are used to generate robust digital terrain models from point cloud data due to higher accuracy and volume of data due to a high sampling rate. LiDAR is able to generate real-time maps with high resolution with fewer computational resources and time needed compared to photogrammetry. Furthermore, the compact size of point cloud data reduces the bandwidth required for sharing data across the drone swarm. The results demonstrate the feasibility of a low-cost decentralized drone swarm architecture with highly efficient and accurate real-time 3D mapping capabilities, enabling faster formation of real-time maps. LOUISIANA

Competition history

• JSHS 2023 Category not listed

Resources

• National JSHS abstract book ↗

Related projects

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