Ho How w to to Detec Detect soft soft falls alls on on de - - PowerPoint PPT Presentation

ho how w to to detec detect soft soft falls alls on on de
SMART_READER_LITE
LIVE PREVIEW

Ho How w to to Detec Detect soft soft falls alls on on de - - PowerPoint PPT Presentation

Ho How w to to Detec Detect soft soft falls alls on on de devices vices some signal processing with knime Prof. Dr. Dominique Genoud Vincent Cuendet master HES-SO, Institute of Information Systems Lausanne, Switzerland


slide-1
SLIDE 1

Ho How w to to Detec Detect soft soft falls alls

  • n
  • n

de devices vices

some signal processing with knime”

  • Prof. Dr. Dominique Genoud

Institute of Information Systems Dominique.Genoud@hevs.ch Techno-Pôle 3 – CH-3960 Sierre Switzerland Vincent Cuendet master HES-SO, Lausanne, Switzerland Julien Torrent FST, Neuchâtel,Switzerland

slide-2
SLIDE 2

Processing flow

slide-3
SLIDE 3

Connected android watches

Moto 360 and LG-G watches

slide-4
SLIDE 4

Use the magnitude of 3D vector

slide-5
SLIDE 5

Falls

t

About 98% detection

slide-6
SLIDE 6

Soft falls

[ms]

1% detection with thresholds

slide-7
SLIDE 7

Signal processing to find patterns

  • Decomposition of a temporal signal:

Fast Fourrier Transform (FFT) In the digital word : Discrete Fourrier Transform (DFT)

slide-8
SLIDE 8

Knime to perform DFT

Reading the samples, and computing the magnitude vector

slide-9
SLIDE 9

Knime to perform DFT

slide-10
SLIDE 10

Knime to perform DFT

slide-11
SLIDE 11

Knime to perform DFT

  • Jwave.jar:

– open source library developed by C. Scheiblich (MIT)

  • The source code and examples are

available there:

– https://github.com/cscheiblich/JWave

  • License free :
slide-12
SLIDE 12

Other polular pattern detection

  • Wavelet transform :

https://en.wikipedia.org/wiki/Wavelet_transform

slide-13
SLIDE 13

Knime to perform Wavelets

  • Implemented in the jwave.jar library:

Haare Legendre Daubechie (db2)

slide-14
SLIDE 14

Knime to perform Wavelets

slide-15
SLIDE 15

Processing flow

slide-16
SLIDE 16

Classification experiments

slide-17
SLIDE 17

Results

Preprocessing Predictor AUC precision FFT 8 coefficients Decision Tree Ensemble 0.84 ±0.03 Mulltilayer Perceptron 0.79 ±0.03 Wavelets Haare Decision Tree Ensemble 0.87 ±0.03 K Nearest Neibourghs 0.75 ±0.04 Daubechie 128 coefficients Decision Tree Ensemble 0.86 ±0.03 K Nearest Neibourghs 0.73 ±0.04

Stratified Subset of 20% of the original data

Full dataset: 500 soft falls and 1500 normal activities

slide-18
SLIDE 18

Scoring and precision of AUC

slide-19
SLIDE 19

How good are the Decision Tree Ensemble

slide-20
SLIDE 20

Processing flow

slide-21
SLIDE 21

Implementation on Android

Jpmml - open source library On existing App

slide-22
SLIDE 22

Enhancements

  • Pmml to java compiler
  • Wrapped nodes
  • Optimize the size of the pmml in memory
slide-23
SLIDE 23

References

  • The example workflows will be available
  • Paper on soft falls
slide-24
SLIDE 24

Questions?

24