Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Improving mmWave based Hand Hygiene Monitoring through Beam Steering and Combining Techniques

Isura Nirmal, Wen Hu, Mahbub Hassan, Elias Aboutanios, Abdelwahed Khamis

TL;DR

BMX addresses the challenge of accurate micro-gesture recognition at extended ranges with commodity mmWave radars by combining transmit beam steering across multiple directions with a deep-learning fusion module. The method constructs multi-view Range-Doppler data per beam, augments data to cover Doppler and multipath variability, and fuses beams through a self-attention transformer followed by CNN-LSTM stages. Across 7,200 WHO-recommended hand-rub gestures from 10 subjects at 1.5 m, BMX achieves up to 91.8% accuracy at boresight using two beams and outperforms state-of-the-art methods by 31–43%, while maintaining robust performance to orientation and distance changes. These results demonstrate BMX’s potential for scalable, privacy-preserving hand hygiene monitoring and motivate extending beam-steering fusion to other mmWave sensing tasks.

Abstract

We introduce BeaMsteerX (BMX), a novel mmWave hand hygiene gesture recognition technique that improves accuracy in longer ranges (1.5m). BMX steers a mmWave beam towards multiple directions around the subject, generating multiple views of the gesture that are then intelligently combined using deep learning to enhance gesture classification. We evaluated BMX using off-the-shelf mmWave radars and collected a total of 7,200 hand hygiene gesture data from 10 subjects performing a six-step hand-rubbing procedure, as recommended by the World Health Organization, using sanitizer, at 1.5m -- over five times longer than in prior works. BMX outperforms state-of-the-art approaches by 31--43% and achieves 91% accuracy at boresight by combining only two beams, demonstrating superior gesture classification in low SNR scenarios. BMX maintained its effectiveness even when the subject was positioned 30 degrees away from the boresight, exhibiting a modest 5% drop in accuracy.

Improving mmWave based Hand Hygiene Monitoring through Beam Steering and Combining Techniques

TL;DR

BMX addresses the challenge of accurate micro-gesture recognition at extended ranges with commodity mmWave radars by combining transmit beam steering across multiple directions with a deep-learning fusion module. The method constructs multi-view Range-Doppler data per beam, augments data to cover Doppler and multipath variability, and fuses beams through a self-attention transformer followed by CNN-LSTM stages. Across 7,200 WHO-recommended hand-rub gestures from 10 subjects at 1.5 m, BMX achieves up to 91.8% accuracy at boresight using two beams and outperforms state-of-the-art methods by 31–43%, while maintaining robust performance to orientation and distance changes. These results demonstrate BMX’s potential for scalable, privacy-preserving hand hygiene monitoring and motivate extending beam-steering fusion to other mmWave sensing tasks.

Abstract

We introduce BeaMsteerX (BMX), a novel mmWave hand hygiene gesture recognition technique that improves accuracy in longer ranges (1.5m). BMX steers a mmWave beam towards multiple directions around the subject, generating multiple views of the gesture that are then intelligently combined using deep learning to enhance gesture classification. We evaluated BMX using off-the-shelf mmWave radars and collected a total of 7,200 hand hygiene gesture data from 10 subjects performing a six-step hand-rubbing procedure, as recommended by the World Health Organization, using sanitizer, at 1.5m -- over five times longer than in prior works. BMX outperforms state-of-the-art approaches by 31--43% and achieves 91% accuracy at boresight by combining only two beams, demonstrating superior gesture classification in low SNR scenarios. BMX maintained its effectiveness even when the subject was positioned 30 degrees away from the boresight, exhibiting a modest 5% drop in accuracy.

Paper Structure

This paper contains 26 sections, 10 equations, 15 figures, 6 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Primer on mmWave Radar
  3. Motivation and Key Idea
  4. BMX Methodology
  5. Transmit beamsteering with COTS mmWave radars
  6. Signal Processing Pipeline
  7. Data Augmentation
  8. Deep Learning Model
  9. Evaluation
  10. Goals, Metrics and Methodology
  11. Hand hygiene monitoring
  12. Data Collection
  13. Deep Model Training Parameters
  14. Metrics and benchmarks
  15. Overall Performance
  16. ...and 11 more sections

Figures (15)

  • Figure 1: Environmental and application considerations can prohibit a user-facing deployment setup (red) assumed by most RF sensing systems. Above is an example of RF-based hand hygiene compliance monitoring. We propose BMX that uses special processing relying on Tx beamforming) to address a more practical yet challenging deployment setup (green).
  • Figure 2: BMX high-level system overview. BMX combines advances in both radar signal construction and deep learning to enable accurate recognition of intricate hand gestures (e.g., hand rub label $G_3$) under challenging deployment setups.
  • Figure 3: Concept of TDM-MIMO and Tx based beamforming
  • Figure 4: Different multi-paths created by different beams.
  • Figure 5: Range profiles for the same gesture observed by two different beams yielding different multipath structures.
  • ...and 10 more figures