Pilot 03 · Vinhomes & Vinmec

Smart Homes & Aging-in-Place

Passive fall detection and behavior monitoring using UWB radar—enabling safe, independent living.

Pilot Status: Planned | Target: Q1 2027 | Scale: 10–15 Vinhomes units, instrumented with edge confidence monitoring

The Gerontology Challenge

Fall Detection Without Cameras

Falls are the leading cause of injury in aging populations. Most fall-detection systems are either wearables (many seniors forget/refuse to wear them) or cameras (privacy invasion). Pilot 03 deploys radar-based, privacy-preserving fall detection with self-aware confidence scoring—detecting both macro-movements (falls) and micro-movements (respiratory anomalies) without invading privacy.

Technical Architecture

Radar-Based Behavioral Sensing

RF Micro-Doppler Signature Classification

Ultra-wideband (UWB) radar sensors running highly optimized, low-power micro-Doppler classification models at the edge:

  • Macro-Movement Tracking — fall velocity signatures, gait patterns, lying-down vs. standing posture differentiation
  • Micro-Movement Resolution — respiration rate extraction (chest wall micro-motion), heart rate inference via cardio-genic micro-Doppler
  • Activity Recognition — walking, sitting, resting, agitated motion differentiation via time-frequency spectral analysis
  • No Privacy Invasion — purely kinematic, no imaging, no facial recognition, radar-only architecture

Edge Confidence & False-Alarm Reduction

On-device uncertainty quantification dramatically reduces false alarms while maintaining genuine alert sensitivity:

  • Fall Confidence Scoring — multi-channel RF correlation detects genuine falls from accidental radar reflections or pets
  • Alarm Severity Tiers — high-confidence falls trigger immediate responder alert; medium-confidence events trigger caregiver notification; low-confidence events logged privately
  • Adaptive Learning — baseline respiration/heart-rate models personalized per occupant to reduce behavioral noise variation
  • Environmental Resilience — RF signatures calibrated for furniture/fixtures per-room, reducing multi-path artifacts
System Specifications
RF Technology: Ultra-Wideband (UWB), 3.1–10.6 GHz, <0.5 m range resolution, 10 Hz frame rate
Coverage: Single unit covers ~10 m² (bedroom, living room), multi-unit mesh for larger spaces
False Alarm Rate: <2% per 24 hours with confidence scoring; <0.5% for high-confidence thresholds
Detection Latency: Fall alert <1 sec post-event, continuous vitals monitoring 1 Hz update rate
Live Demonstration

Micro-Doppler Radar Kinematics

Experience how camera-less fall detection operates at the edge. Select an activity profile below to simulate human movement inside a radar transceiver zone. The engine computes a Short-Time Fourier Transform (STFT) approximation, rendering a live spectral velocity map showing how RF signatures change across different behavioral patterns.

Target Activity Profile
DSP Core Status
⚡ STREAMING NOMINAL
Computed Velocity
1.2 m/s
Spectral Power Bandwidth
140 Hz
Autonomy Status
ROOM SECURE
+Doppler (Approaching)
-Doppler (Receding)
Timeline (Real-Time Scrolling Window)
Signal Magnitude (dB):
Max Peak
Advanced Visualization

3D Beamforming & Target Localization

Experience sophisticated MIMO radar signal processing: real-time 3D target localization with adaptive beamforming. The antenna array dynamically phases its transmission to focus energy toward targets while rejecting interference. Watch the heatmap reveal signal strength and the confidence ellipsoid show localization uncertainty.

6.5 GHz
Target Distance
5.2 m
Azimuth Angle
45°
SNR (dB)
18.5
Localization Error
±0.3 m
🔵 Blue: Antenna Array Elements 🟢 Green: Target Position 🌡️ Heatmap: RF Signal Strength ⭕ Ellipsoid: Confidence Region
Deployment Model

Pilot Cohort & Measurement Plan

Cohort Characteristics
  • • 10–15 senior residents (65–85 years old)
  • • Mixed mobility levels (independent to mobility-limited)
  • • Diversified living arrangements (1BR, 2BR, shared spaces)
  • • Baseline health screening + ongoing caregiver engagement
Duration: Q1 2027 – Q4 2027 (continuous monitoring)
Key Measurements
  • • Fall detection sensitivity & false-alarm rate per occupant
  • • Respiration/heart-rate accuracy vs. reference wearables
  • • Caregiver response time & engagement metrics
  • • Occupant privacy & social acceptance survey data
Analysis: Monthly reports, Q4 final outcomes paper
Core Research Frontiers

This pilot draws on V-SENSE's three core engineering frontiers:

Frontier 01
Self-Aware CMOS
Frontier 02
Neuromorphic Edge
Frontier 03
Reconfigurable RF