1: Université Pierre et Marie Curie – Paris 6

Laboratoire d’Électronique et d’Électromagnétisme - L2E

2: Chaire Ebiomed: Université de Technologie Compiègne

BioMécanique et BioIngénierie – BMBI (UMR 7338)

Détection de chute en utilisant

des techniques Radar

Ting ZHANG1,2, Julien SARRAZIN1, Guido VALERIO1, Dan ISTRATE2

Evry, November 23th, 2016

Colloque PARAChute 2016

Outline

Context

Objective

Proposed solution

Experimental results

Conclusions and perspectives

2

Context

3

Elderly people fall: 2 millions per year in

France (leading to 10,000 deaths)

For people >80 years old: 50% falls once

per year

Quick medical assistance may save life

Life expectancy in France

Female

Male

Problem of slow falls

Objective

4

Monitoring emergency situations

Active Motionless

Physiological rythm:

Breathing and HeartbeatingTracking

)(tx

d

Doppler radar at 60 GHz

Wavelength: 5 mm

Person keeping still: d is a constant

)(

)(442cos)( t

txdftAtR res

r

Solution: Doppler radar

5

ftAtT e 2cos)(

Physiological movements

Breathing Heartbeating

)sin( rrr tm )sin( hhh tm +

)(tx

I/Q quadrature demodulation

6

RXLNA

Coupler 3dB (90°)

Q

IAmp

Amp

BPF

BPF

Saving data

and signal

processing

)(

)(442cos)( t

txdftAtR res

r

Baseband: BI

Complex demodulation

LO

)sin()sin(exp)exp(

)(

44

hh

m

rr

m

QI

ttii

jBBtB

hr

Baseband: BQ Breathting Heartbeating

7

Baseband signal

mr = 1.0 mm, mh = 0.1 mm, fr = 18 bpm, fh=72 bpm

In presence of random body movements

Heartbeating

Random body

movements

Breathing

FFT

Fundamental term of breathing

Heartbeating ???

8

Direct peak detection in the spectrum

......

Harmonics, Intermodulation

Solution: Optimization problem

Complex received signal

Determination of all unknown parameters

Optimization algorithms: Particle Swarm Optimization (PSO), Genetic Algorithm (GA)

Cost function:

9

hrhrhr ffmm

)2sin()2sin(exp)exp()(44

hh

m

rr

mcal tftfiitB hr

fr (bpm) fh (bpm) mr (mm) mh (mm)

At rest lb 12 54 0 0

ub 25 100 6.0 1.0

After sport lb 25 100 0 0

ub 60 180 6.0 1.0

2

2

)(

)()(

mes

calmes

tB

tBtBF

Doppler Radar

10

Experimental results

Experimental setup

11

Multiplier ×4, LO

55GHz

Amplifier

90° coupler

IF 5GHz

Down-Converters

Transmitter: RF 60 GHz

Receiver

Low Noise Amplifier

1m in front of antennas

Sensors: reference

LO: 13.75 GHz

Measured baseband signal

12

Each measurement: 30 seconds

B_I

B_Q

Optimization results: breathing

13

The frequency is correctly estimated

Variation of the amplitude of movement is

detectable

Optimization results

Respiratory belts

Optimization results: heartbeating

14

Simu SNR 10dB

Simu SNR 9dB

Simu SNR 6dB

Exp back

Exp front

5 people, front and back, 80 mesurements for each scenario

SNR of the actual system: between 6 and 9 dB

Conclusion and Perspective

15

Conclusion

60 GHz Doppler radar with optimization algorithm enables

autonomous vital signs estimation

We can detect not only the frequencies, but also the displacements

introduced by the physiological movements

Perspective

Improving the system for a better accuracy of estimation

Use wavelet transform

Thank you

16