Fully Adaptive FMCW Radar System for Gesture Recognition

This repository contains the code for the paper entitled "Human-Centered Fully Adaptive Radar for Gesture Recognition in Smart Environments" which was published in IEEE Transactions on Human-Machine Systems. The radar used in this study is the Infineon BGT60TR13C FMCW Radar.

Directory Structure

• configs: Configuration files for the radar system. Available modes are: simo, light, trigger as described in the paper.
• data_processing: Scripts for data post-processing which generate range-profiles and angle-profiles and micro-Doppler spectrograms. Also contains scripts for data labeling (if recorded with main_recorder.py).
• dnn: Deep learning models for training gesture recognition models.

Installation

# Clone the repository
git clone https://github.com/ekurtgl/Fully_Adaptive_Radar-Infineon_BGT60TR13C.git

# Navigate to the repository
cd Fully_Adaptive_Radar-Infineon_BGT60TR13C

# Create a virtual environment (optional)
python -m venv venv

# Activate the virtual environment
venv\Scripts\activate

# Setup Radar
# Follow setup_radar.ipynb to setup the radar

# Install dependencies
pip install -r requirements.txt

Multi-Modal (Radar + Camera (Optional)) Data Recording (No Inference)

To record data with main_recorder.py:

# change the operating mode in line #15 of main_recorder.py
# change the data saving directory in line #18 of main_recorder.py
python main_recorder.py

Model Training

dnn/DNN*.ipynb files are for training the deep learning models.

Real-Time Fully Adaptive Inference

main_realtime.py file is for real-time fully adaptive inference. data_path, trigger_model and recognition_model should be set to the directory containing the trained model weights.

Cite as:

Bibtex:
@ARTICLE{11126867,
  author={Kurtoğlu, Emre and Gurbuz, Sevgi Z.},
  journal={IEEE Transactions on Human-Machine Systems}, 
  title={Human-Centered Fully Adaptive Radar for Gesture Recognition in Smart Environments}, 
  year={2025},
  volume={55},
  number={5},
  pages={695-706},
  keywords={Radar;Radio frequency;Gesture recognition;Accuracy;Sensors;Radar tracking;Human activity recognition;MIMO;Human computer interaction;Frequency-modulated continuous wave (FMCW);hand gesture recognition;human activity recognition;MIMO;radar},
  doi={10.1109/THMS.2025.3591369}}