Motion artifacts reduction in cardiac pulse signal acquired from video imaging
PPG_Bio_Recog
1. A study based on usage of
PhotoPlethysmoGraphic (PPG) Signals
in the Biometric Recognition
Presented by
Bidhan Barai
2. Definition
●
A photoplethysmogram (PPG) is an optically obtained plethysmogram, a
volumetric measurement of an organ.
●
A PPG is often obtained by illuminating the skin and measures changes in
light absorption
●
The change in volume caused by the pressure pulse is detected by
illuminating the skin with the light from a light-emitting diode (LED) and
then measuring the amount of light either transmitted or reflected to a
photodiode.
3. Why PPG Signal?
●
The characteristics of human body that are used in Biometric
Recognition of Human: Fingerprint, Face, Voice, Retina/Iris, Lip
Movement, Gait motion. However, fingerprint can be recreated in
latex, face recognition can be fooled by a photo, voice can be imitated.
●
Compared with other biometric approaches, PPG technique has
several distinct advantages including low development cost, easy to use
without any complicated procedure or special skill, and conveniently
accessible to various sites of human body, such as finger, ear lobe,
wrist or fore head.
5. Acquisition Of PPG Signal
●
Basics of Color: When light hits an object various frequencies are
absorbed and transmitted through the object while other
frequencies are reflected.
6. Acquisition Of PPG Signal
●
NJL5501R is a Reflective type Optical Sensor.
7. Acquisition Of PPG Signal
●
Reflected and Transmitted light gives some valuable information
about the object. The PPG signals are formed by following this
basic theory of light.
8. Acquisition Of PPG Signal
●
DCM03 Optical Sensor (Transmission Type) is used for this
purpose.
9. Acquisition Of PPG Signal
●
Oxygenated hemoglobin (HbO2) absorbs more infrared light and allows
more red light to pass through. Deoxygenated (or reduced) hemoglobin
(Hb) absorbs more red light and allows more infrared light to pass
through. Red light is in the 600-750 nm wavelength light band. Infrared
light is in the 850-1000 nm wavelength light band.
11. Pre-Processing
●
Elimination of Noise:
1> A Low Transition FIR filter with N=200 points 10 Hz
cutting frequency is used.
2> A upper exemplification (Up Sampling) process with
exemplification factor 4 is performed to increase frequency
from 16.5 Hz to 66 Hz.
4
Low Pass Filter
(Gain: 4)
16.5 Hz f c=π/4 66 Hz
12. Time domain Feature Extraction
●
By the use of the original PPG signal and its two derivatives (1st
and 2nd
order derivatives)
40 characteristic features were calculated. such as x (systolic peak), y (diastolic peak), z
(dicrotic notch), tpi (pulse interval), tpp (peak to peak), y/x (augmentation index), (x -
y)/x (alternative augmentation index), z/x|(y - z)|/x, t1 (systolic peak time), t2 (dicrotic
notch time), t3 (diastolic peak time), ΔT (time between systolic and diastolic peaks),
width (the pulse width with semi-height of the systolic peak), A2/A1 (inflection point area
ratio-IPA), t1/x (systolic peak output curve), y/(tpi - t3) (diastolic peak downward curve),
t1/tpp, t2/tpp, t3/tpp, ΔT/tpp, ta1, tb1, te1, tf1, b2/a2, e2/a2, (b2+c2)/a2, ta2, tb2,
ta1/tpp, tb1/tpp, te1/tpp, tf1/tpp, ta2/tpp, tb2/tpp, (ta1 - ta2)/tpp, (tb1 - tb2)/tpp, (te1
- t2)/tpp, (tf1 - t3)/tpp.
13. Time domain Feature Extraction
●
1st
and 2nd
Derivatives of PPG Signal:
2nd
Order Derivative
1st
Order Derivative
15. Classification
●
The most popular method which makes classification by using the
distance between characteristic points of Train PPG Signals and Test
PPG Signal is the k-nn (k-nearest neighbour) algorithm.
●
The k-nn Algorithm:
1> “k”value is total number of person.
2> The distances between the characteristic property points whose
class labels are not known and whose class labels are known are
claculated. Generally Euclid Distance is used.
16. Classification
●
The k-nn Algorithm:
3> The distances are put in order. The nearest k-class label
with smallest distance is defined.
4> The class label with the majority among other k-class
labels is defined. This defined label is assigned as the result of the
unknown class.
5> After this cluster centres are updated.
17. Classification
●
Cross Validation method should be used to measure the
classification success in an unambigous manner.
●
The types of cross validation methods are suggested in the
literature:
1> Random Exemplification.
2> K-pieced.
3> Leave-One-Out.
18. Other Usage of PPG Signal
●
The most popular use of PPG signal is to measure the Heart
Rate. The device is called Pulse Oximeter.
20. References
[1] A. Reşit Kavsaoğlu , Kemal Polat , M. Recep Bozkurt - “A novel
feature ranking algorithm for biometric recognition with PPG signals”.
Computers in Biology and Medicine, ELSEVIER.
[2] Petros Spachos, Jiexin Gao and Dimitrios Hatzinakos - “FEASIBILITY
STUDY OF PHOTOPLETHYSMOGRAPHIC SIGNALS FOR
BIOMETRIC IDENTIFICATION”. IEEE.
[3] A. Reşit Kavsaoğlu , Kemal Polat , M. Recep Bozkurt, Hariharan
Muthusamy - “Feature Extraetion for Biometrie Reeognition with
Photoplethysmography Signals”. IEEE