Single mmWave Radar-Based Pose Detection for Non-Intrusive Patient Monitoring

A Three-Stage Pipeline to predict and depict a person's pose

Repository

This repository contains the following folders and code files

Folders

1. 01_raw_bin - Containing the raw binary files collected from the mmWave Radar
2. 02_parsed - Folder for the parsed binary files
3. 03_occupancy_grid - folder for the generated occupancy grid

Code Files

1. data_process.ipynb - To proces the raw binary files into readable MS Excel format, read the data, and generate occupancy grids
2. model_training.ipynb - For model training
3. pose_depiction.ipynb - For prediction and depicting the pose

mmWave Radar Details

IWRL6432BOOST Evaluation Module

Radar Configuration

• Demo used - IWRL6432 Out of Box Demo
• Chirp configuration - PresenceDetect.cfg (default configuration provided in the radar toolbox)
  • The configuration is altered to retrieve more point cloud data
• Within the visualizer, a save UART option can be checked, allowing the point cloud to be collected
  • The data points are saved as a .bin file after every 100 frames, and through exiting the tool, remaining frames will be discarded

Usage

1. Install dependencies from a requirements.txt file

pip install -r requirements.txt

2. Run each .ipynb files in sequence