Target Classification on Edge Using mmWave Radar - A Novel Algorithm and Real-Time Implementation

Watch a 21-minute conference talk from tinyML Asia 2023 exploring an innovative approach to target classification using mmWave radar technology on edge devices. Learn about a novel algorithm developed for TI's IWRL6432 radar-on-chip that enables real-time multiple-target classification in cluttered environments. Discover how the system processes radar signals through multiple stages, from generating radar cubes to creating µ-Doppler spectrograms, and implements a unique 1D-CNN classification model optimized for edge computing. Understand the technical challenges overcome in developing a solution that can simultaneously track and classify multiple targets, distinguishing between human and non-human objects with limited computational resources. Gain insights into the extensive training process involving over 125,000 frames of data captured across 310 different scenarios, featuring various human and non-human targets. Follow along as the speaker details the complete implementation process, from hardware components and SDK integration to live demonstrations of the system's capabilities.