Indian Institute of Information Technology - Allahabad - Disputationes
Indian Institute of Information Technology - Allahabad - Disputationes Indian Institute of Information Technology - Allahabad - Disputationes
2nd Disputations on Future Technologies for Health Care Allahabad (India), 18-20 September 2011 Diagnosis Through Gait Oscillations Pavan Chakraborty, Soumik Mondal, Anup Nandy, Saman Shahid and G. C. Nandi. Indian Institute of Information Technology, Allahabad, Deoghat, Jhalwa, Allahabad – 211 012 India E-mail: pavan@iiita.ac.in Indian Institute of Information Technology - Allahabad
- Page 2 and 3: Diagnosis Through Gait Oscillations
- Page 4 and 5: Locomotive Balance Static balance s
- Page 6 and 7: Walking: A common yet complex proce
- Page 8 and 9: Indian Institute of Information Tec
- Page 10 and 11: Indian Institute of Information Tec
- Page 12 and 13: Subject with reflective markers for
- Page 14 and 15: Objective “The main objective of
- Page 16 and 17: θ Potentiometer Voltage Calibrated
- Page 18 and 19: The actual knee angle oscillation o
- Page 20 and 21: Indian Institute of Information Tec
- Page 22 and 23: Indian Institute of Information Tec
- Page 24 and 25: Figure 6. Gait oscillations from 8
- Page 26 and 27: Sensor and Interface Kit Specificat
- Page 28 and 29: Analysis of human gait oscillations
- Page 30 and 31: Gait Patterns Obtained… Gait patt
- Page 32 and 33: Correlation and coupling of signifi
- Page 34 and 35: Coupling Between Elbows The above f
- Page 36 and 37: Coupling Between Shoulder and Elbow
- Page 38 and 39: Indian Institute of Information Tec
- Page 40 and 41: List of Publications • Pavan Chak
- Page 42 and 43: Indian Institute of Information Tec
- Page 44 and 45: Indian Institute of Information Tec
- Page 46 and 47: Conclusion • A biometric suit IGO
- Page 48 and 49: Pavan Chakraborty, Anup Nandy and G
- Page 50 and 51: Indian Institute of Information Tec
2nd Disputations on<br />
Future Technologies for Health<br />
Care<br />
<strong>Allahabad</strong> (India), 18-20 September 2011<br />
Diagnosis<br />
Through<br />
Gait Oscillations<br />
Pavan Chakraborty, Soumik<br />
Mondal, Anup Nandy, Saman<br />
Shahid and G. C. Nandi.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong>,<br />
Deoghat, Jhalwa, <strong>Allahabad</strong> – 211 012 India<br />
E-mail: pavan@iiita.ac.in<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Diagnosis Through Gait Oscillations<br />
Learning to walk is a daunting<br />
task for a human baby.<br />
It takes close to a year for a<br />
human baby to stand on its two<br />
legs, balance and then learn to<br />
walk.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Locomotive<br />
Balance<br />
Static balance slower<br />
locomotion.<br />
Dynamic balance Faster locomotion.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Our biped locomotion<br />
was learned on Trees<br />
Orang-utan study suggests that upright<br />
walking may have started in the trees.<br />
--- Branch walking easier on biped,<br />
holding the branch above with hands.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Walking: A common yet<br />
complex process<br />
•Human walking is<br />
accomplished with a strategy<br />
called the double pendulum<br />
•During forward motion, the leg<br />
that leaves the ground swings<br />
forward from the hip<br />
•Leg strikes the ground with<br />
the heel and rolls through to<br />
the toe in a motion<br />
•Motion <strong>of</strong> the two legs is<br />
coordinated so that one foot or<br />
the other is always in contact<br />
with the ground<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Foot<br />
Pressure<br />
Platform<br />
Animations<br />
Contact points <strong>of</strong> the sole <strong>of</strong> the foot during<br />
normal weight bearing.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Subject with reflective markers for motion analysis. Also<br />
seen are the electrodes for the telemetry EMG unit<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Use <strong>of</strong> Sensor Network<br />
To Understand<br />
The Human Gait<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Objective<br />
“The main objective <strong>of</strong> this research tends to develop a low<br />
cost, portable, non-invasive wearable sensor based<br />
biometric suit which could be applied to numerous different<br />
applications.” Like,<br />
• Human Identification<br />
• Human Robot Interaction<br />
• Gesture Classification<br />
• Human Computer Interaction<br />
• Diagnostic Gait Signature Detection<br />
• Find the effect on gait with different age, body weight and<br />
different diseases.<br />
• Analysis <strong>of</strong> abnormal gait signature etc.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
The Degrees <strong>of</strong> Freedom<br />
3<br />
3 3<br />
1 1<br />
1<br />
1 1<br />
3 3<br />
1<br />
1<br />
2<br />
2<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
θ<br />
Potentiometer<br />
Voltage Calibrated to Angle θ<br />
θ<br />
t<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
θ ( t)<br />
= −65.71<br />
V ( t)<br />
+ 122.9<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
The actual knee<br />
angle oscillation<br />
<strong>of</strong> a healthy<br />
human being<br />
measured using<br />
a potentiometer<br />
circuit. The top<br />
panel shows the<br />
noisy data over<br />
a multiple gait<br />
oscillations.<br />
Averaging multiple gaits and convolving it with a 3 value running<br />
average in 2 iterations removes the noise (shown in the lower<br />
panels).<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
8 Analog Inputs<br />
AVR Dev. Board<br />
ACD, Multiplexing &<br />
Parallel to Serial<br />
conversion.<br />
ATMEGA 32<br />
RF Module Tx<br />
RF Module Rx<br />
Max 232<br />
RS – 323 Port<br />
Serial Port<br />
S<strong>of</strong>tware<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
The HGOD Suit consisting <strong>of</strong> 8 potentiometer sensors<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Figure 6. Gait oscillations from 8 joints <strong>of</strong> the body. From the top (left and<br />
right): Left and right Shoulders, left and right elbows, left and right hips and<br />
left and write knees. The dotted line is a simple sine curve fit to the shoulder<br />
elbow and hip data. The knee oscillation is complex; therefore it has not been<br />
fitted.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Design…<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Sensor and Interface<br />
Kit Specification<br />
• Phidget Rotation sensor (1109) which is being<br />
integrated with Phidget interface kit (1018).<br />
• Power supply voltage varies from 3.5VDC to 5 VDC with<br />
10kΏ output impedance<br />
• The rotation sensor has been opted for 0 to 300 degree<br />
resolution<br />
• Measurement <strong>of</strong> analog value from the rotation sensor<br />
and produces the digital counts as output between 0 to<br />
1000 ranges<br />
• Analog Input update rate is 65 samples/sec<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Calibration Curve for Sensor<br />
• Phidget interface kit has<br />
Internal 10 bit ADC<br />
circuit<br />
• Equation depicting linear<br />
relationship between the<br />
observed data and<br />
calibrated data with least<br />
square fitting is : θ = 3.3335 x<br />
count (±0.020161)<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Analysis <strong>of</strong> human gait<br />
oscillations<br />
• Initially, for each gait pattern <strong>of</strong> respective joint a<br />
zero correction has already been done by collecting<br />
α 0 represented in the form <strong>of</strong> digital counts.<br />
• The initial digital counts are subtracted from the<br />
current digital count which is being interpolated by<br />
joint angle values in degree for each oscillation.<br />
• The movement <strong>of</strong> each oscillation for a particular<br />
joint is manipulated in terms <strong>of</strong> degree which is<br />
being calculated by<br />
desired angle (degree)=(current digital count-initial<br />
count)x 300/1000.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Gait Patterns Obtained<br />
Gait pattern <strong>of</strong> both elbow joints<br />
Gait pattern <strong>of</strong> both knee joints<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Gait Patterns Obtained…<br />
Gait pattern <strong>of</strong> both hip joints<br />
Gait pattern <strong>of</strong> both shoulder joints<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Observations<br />
• Each gait oscillation consists <strong>of</strong> both swing and stance<br />
phase respectively.<br />
• The generated pattern for both left limb and right limb<br />
implies the variation <strong>of</strong> calibrated angle values over the<br />
period <strong>of</strong> oscillation<br />
• The period <strong>of</strong> each oscillation for a particular pattern can<br />
be calculated by<br />
Total no <strong>of</strong> samples in a particular oscillation 64(approx)<br />
T =<br />
=<br />
= 1sec(approx)<br />
sampling rate<br />
65<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Correlation and coupling <strong>of</strong><br />
significant joints<br />
• Correlation and coupling <strong>of</strong> those joint<br />
oscillations, we have compared them with an<br />
oscillation equation <strong>of</strong> the form:<br />
X<br />
( t)<br />
= asin(<br />
ω 1t<br />
+ ϕ)<br />
Y<br />
( t)<br />
= bsin(<br />
ω 2t)<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Coupling Observations<br />
It has been observe from the figures that phase difference φ from the fitting<br />
between two gait oscillation is |φ|=π and<br />
ω1 = ω 2 = ω = 2π<br />
/ T = 2π<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Coupling Between Elbows<br />
The above figure shows an interesting oscillation <strong>of</strong> an envelope in the shape<br />
<strong>of</strong> an ‘L’. This indicates when one elbow oscillates the other is practically<br />
static. This happens because <strong>of</strong> the elbow locking in the reverse cycle.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Coupling Between Shoulder<br />
and Hips<br />
It is being noted from the above figures that the coupling between both<br />
shoulder and hip oscillations tends to an elliptical curve where phase<br />
difference |φ|= 5π/4 .<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Coupling Between Shoulder<br />
and Elbows<br />
From the above figures it has been noticed that the movement <strong>of</strong><br />
shoulder oscillation arises in both directions where as the oscillation <strong>of</strong><br />
elbow joint belongs to in single direction<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Creation <strong>of</strong> a Gait Data Base<br />
Gait Oscillations at Different Speed.<br />
With Different Carrying Weights.<br />
Other Biometric <strong>Information</strong> for Correlations:<br />
Male/Female,<br />
Weight<br />
Height<br />
Age<br />
Body Shape – Front and side view photo<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
List <strong>of</strong> Publications<br />
• Pavan Chakraborty, Anup Nandy and G. C. Nandi, 2007.<br />
• "Use <strong>of</strong> Sensor Network to Understand the Human Gait"<br />
• An Invited Oral Presentation WCSN-2007 International<br />
Conference held at IIIT-<strong>Allahabad</strong>.<br />
• Pavan Chakraborty, G. C. Nandi and Anup Nandy, 2008.<br />
• “Sensing the Human Locomotion”<br />
• An Invited Oral Presentation at the Multi Discipline Expert<br />
Panel Meeting held at ALIMCO Kanpur on January 30 and 31,<br />
2008.<br />
• Akshay K Singh, Ajit D Dhiwal, Pavan Chakraborty and G C<br />
Nandi. 2008. “Designing A Full Body Human Computer<br />
Interaction Device” Proceedings <strong>of</strong> the National Conference<br />
CSI-RDHS 2008 Research and Development in Hardware and<br />
Systems by Computer Society <strong>of</strong> India).<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
Continued………
List <strong>of</strong> Publications<br />
• “A framework for synthesis <strong>of</strong> human gait oscillation using intelligent gait<br />
oscillation detector (IGOD)” In: IC3 2010. CCIS, vol. 94, pp. 340–349.<br />
Springer, Heidelberg, 2010.<br />
• “Gait Based Personal Recognition System Using Rotation Sensor,”<br />
Submitted in IEEE Trans. On System, Man Cybernetics part: B (Under<br />
Review).<br />
• “Modeling a Central Pattern Generator to generate Biped Locomotion <strong>of</strong> a<br />
Bipedal Robot using Rayleigh Oscillators” – In the proceedings <strong>of</strong> S. Aluru<br />
et al. (Eds.): IC3 2011, CCIS 168, pp. 289–300, Springer-Verlag Berlin<br />
Heidelberg, 2011. (In press)<br />
• “A Central Pattern Generator based Nonlinear Controller to Simulate Biped<br />
Locomotion with a Stable Human Gait Oscillation” – In International Journal<br />
<strong>of</strong> Robotics and Automation (IJRA), vol. 2, issue 2, pp. 93-106, 2011.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Conclusion<br />
• A biometric suit IGOD was build to make simultaneously measurement <strong>of</strong> the 8<br />
major joint oscillations that plays a crucial role during human biped locomotion.<br />
• The mechanical structure <strong>of</strong> IGOD was completed, fine-tuning and calibrations was<br />
done.<br />
• The concept <strong>of</strong> IGOD is extremely simple with significant biometric and robotics<br />
application.<br />
• Analysis <strong>of</strong> the data using correlation coupling between pair <strong>of</strong> left and right limbs<br />
will be an important diagnostics <strong>of</strong> the bio information.<br />
• The gait recognition study has been made using the IGOD data from various<br />
people.<br />
• The TS-LDA developed for the purpose provided promising results.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Conclusion…<br />
• The results were doubly checked with the well established ANN algorithm.<br />
• Interestingly in a different work on gesture based recognition TS-LDA did not<br />
work.<br />
• This made us analyze the capability <strong>of</strong> TS-LDA which is well tuned to detect small<br />
variations which is normally seen in gait recognition; it does not work well for large<br />
variation as seen in gesture recognition.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Pavan Chakraborty, Anup Nandy and G. C. Nandi, 2007.<br />
"Use <strong>of</strong> Sensor Network to Understand the Human Gait"<br />
An Invited Oral Presentation WCSN-2007 International Conference held<br />
at IIIT-<strong>Allahabad</strong>.<br />
Pavan Chakraborty, G. C. Nandi and Anup Nandy, 2008.<br />
“Sensing the Human Locomotion”<br />
An Invited Oral Presentation at the Multi Discipline Expert Panel Meeting<br />
held at ALIMCO Kanpur on January 30 and 31, 2008.<br />
Akshay K Singh, Ajit D Dhiwal, Pavan Chakraborty and G C Nandi.<br />
2008. “Designing A Full Body Human Computer Interaction Device”<br />
Proceedings <strong>of</strong> the National Conference CSI-RDHS 2008 Research and<br />
Development in Hardware and Systems by Computer Society <strong>of</strong> India).<br />
Pavan Chakraborty, Anup Nandy and G. C. Nandi,<br />
"HGOD A Multi Sensor Network to analyze Human Gait and its stability“<br />
Submitted to The Proceeding IEEE Conference.<br />
Raghvendra Jain, Pavan Chakraborty, and G. C. Nandi<br />
“Energy budgeting for a Humanoid Robot using physics properties”<br />
Raghvendra Jain, K.V.N. Pavan, Pavan Chakraborty, and G. C. Nandi,<br />
“Full body Human Robot real Time Gesture based interaction”<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
The Degrees <strong>of</strong> Freedom<br />
3<br />
3 3<br />
1 1<br />
1<br />
1 1<br />
3 3<br />
1<br />
1<br />
2<br />
2<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Statically stable walking: The projection <strong>of</strong> the robot’s<br />
center <strong>of</strong> gravity always lies within the footprint or the area<br />
between the two footprints <strong>of</strong> the robot.<br />
The locomotion is quite easier to perform but is very slow.<br />
Dynamic walking: The robot’s center <strong>of</strong> gravity outside the<br />
footprints. Corrections must be done at all times to maintain<br />
the robot on his feet.<br />
To achieve stability, it is mandatory to know about the robot’s<br />
dynamics (speed and inertia <strong>of</strong> each <strong>of</strong> its parts).<br />
This method is without a doubt the more efficient in<br />
terms <strong>of</strong> velocity but is very tricky to implement.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Passive walking: No actuators are used and the robot uses its<br />
own weight and dynamics to walk down a slope without external<br />
input or control.<br />
The system has a stable limit cycle and the robots are even<br />
able to deal with small perturbations. We can also mention the<br />
concept <strong>of</strong> passive dynamic walkers. In this case small<br />
power sources on their ankles and/or hips are included to<br />
simulate gravity,<br />
which allows the robots to walk on a level ground without<br />
control servos.<br />
This approach is very important since it allows to learn many<br />
features <strong>of</strong> robot’s dynamics that are also useful for other kind<br />
<strong>of</strong> locomotion [5, 4].<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Discussion<br />
• According to our research works we can conclude that for distance based human<br />
identification using gait not only depends upon the leg feature. Here hand<br />
movement is also an important factor.<br />
• Both the methods gives us 100% accuracy.<br />
• This system can not be use for surveillance security purpose.<br />
• This can be used in personal identification purpose where walker full cooperation is<br />
needed.<br />
• In this system mainly we are trying to find out what are the basic features to<br />
recognize the different human walking for identification and research.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
5
Future Work<br />
• Interaction between humanoid robot and human being (Imitation<br />
Learning).<br />
• Classification <strong>of</strong> Gestures for Human Robot Interaction.<br />
• Tuning up the mechanical design and making IGOD wires free will<br />
improve the freedom, flexibility and movability <strong>of</strong> the subject wearing<br />
IGOD.<br />
• Use the suit for medical applications.<br />
• We are also trying to train our system by the suit data and test the<br />
system by image data for giving the freedom <strong>of</strong> walker and use this<br />
system as a gait based human identifier.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
5
List <strong>of</strong> My Publications…<br />
• “Classification <strong>of</strong> <strong>Indian</strong> Sign Language In Real Time” - In the proceedings <strong>of</strong> International Journal<br />
on Computer Engineering and <strong>Information</strong> <strong>Technology</strong> (IJCEIT), Vol. 10, No.15, pp. 52-57, January<br />
- February 2010.<br />
• “Recognition <strong>of</strong> Isolated <strong>Indian</strong> Sign Language gesture in Real Time” – In the proceeding <strong>of</strong><br />
Springer LNCS-CCIS 70, pp. 102-107, March 2010.<br />
• “Recognizing & Interpreting <strong>Indian</strong> Sign Language Gesture for Human Robot Interaction” - In the<br />
proceeding <strong>of</strong> ICCCT’10, IEEE Xplore Digital Library, pp. 712-717, September 2010.<br />
• “Gesture based imitation learning for Human Robot Interaction” – In the proceeding <strong>of</strong> ICACCN’11,<br />
pp. 287-292, 2011.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
5
Pavan Chakraborty<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong>.<br />
A Project funded by DST,<br />
P.I. : Pr<strong>of</strong>. M.D. Tiwari, Director, IIIT-<strong>Allahabad</strong><br />
The Working Team: Pr<strong>of</strong> G. C. Nandi (Head Robotics and AI Lab),<br />
Mr. Anup Nandi and Advitiya Saxena (Research Scholars) & Myself.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
S5<br />
S6<br />
S7<br />
S8<br />
S3<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Classification <strong>of</strong> eight regions over<br />
a gait cycle<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Gait oscillation with noise Vs gait<br />
oscillation with filtering noise<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
AVR 40 Pin Development board<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
RF Module<br />
• An RF modulator<br />
(for radio<br />
frequency<br />
modulator) is a<br />
device that takes a<br />
baseband input<br />
• signal and outputs<br />
a radio frequencymodulated<br />
signal.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Hierarchical<br />
SENSE PLAN ACT<br />
Reactive<br />
Hybrid deliberative / reactive<br />
PLAN<br />
PLAN<br />
SENSE<br />
ACT<br />
SENSE<br />
ACT<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
2nd Disputations on<br />
Future Technologies for Health Care<br />
<strong>Allahabad</strong> (India), 18-20 September 2011<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
· Frames 1 - 3: The robot's momentum causes the robot to rise on its standing leg and<br />
a motor moves the swinging leg into position<br />
· Frame 3: The stretch sensor <strong>of</strong> the swinging leg is activated, which triggers the knee<br />
joint to straighten<br />
· Frames 3 - 6: The robot falls forward naturally, with no motor functions being used,<br />
and catches itself on the next standing leg<br />
· Frame 6: As the swinging leg touches the ground, the ground contact sensor in the<br />
foot triggers the hip extensor and the knee joint <strong>of</strong> the standing leg and the hip and knee<br />
joints <strong>of</strong> the swinging leg to swap roles<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Activation<br />
<strong>of</strong> the calf<br />
muscle<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
M.R. Damper<br />
A typical Magneto-Rheological<br />
(MR) fluid consists <strong>of</strong> 20-40% by<br />
volume <strong>of</strong> relatively pure, 3-10<br />
micron diameter iron particles,<br />
suspended in a carrier liquid such<br />
as mineral oil, synthetic oil, water<br />
or glycol.<br />
Iron particles in suspension align and develop a yield strength<br />
in the presence <strong>of</strong> a magnetic field. The change from a freeflowing<br />
liquid to a semi-solid when a magnetic field is applied is<br />
rapid and reversible.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
11.5mm<br />
35mm<br />
Bearings<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
11.5mm<br />
35mm<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
Bearings
Joint surfaces move with respect to one another by<br />
simultaneously (1) rolling, (2) gliding, and (3) spinning.<br />
When the concave surface is<br />
fixed and the convex surface<br />
moves on it, the convex surface<br />
rolls and glides in opposite<br />
directions.<br />
When the convex surface is fixed<br />
and the concave surface moves<br />
on it, the concave surface rolls<br />
and glides in the same direction.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Hardware Circuit<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
T<br />
ϕ<br />
Assumptions:<br />
•The period <strong>of</strong> the hip cycle is the same as that <strong>of</strong> the knee.<br />
•The phase difference between the hip and knee oscillation<br />
is constant.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Measurement <strong>of</strong> Period T <strong>of</strong><br />
the Hip Oscillation<br />
• The period <strong>of</strong> oscillation could be determined from the<br />
gap between 2 maxima <strong>of</strong> consecutive oscillations<br />
θi<br />
V i<br />
by the POT<br />
V i<br />
V 0<br />
t<br />
pj−1<br />
t<br />
pj<br />
T<br />
j<br />
= t − t<br />
pj pj<br />
−1<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Measurement <strong>of</strong> Period T <strong>of</strong><br />
the Hip Oscillation<br />
V = ∑ j<br />
V i<br />
, VT = ∑ j<br />
Vi<br />
⋅ti,<br />
i<br />
i<br />
for<br />
V i<br />
> V 0<br />
t<br />
pj<br />
=<br />
VT<br />
V<br />
= ∑V<br />
⋅t<br />
i<br />
i<br />
i<br />
∑<br />
i<br />
V<br />
i<br />
T j<br />
T<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
j<br />
= t − t<br />
pj pj<br />
The period , is the time gap between consecutive<br />
peaks j and j-1.<br />
T j<br />
⇒ T 0<br />
−1
Check for Threshold<br />
T −<br />
j<br />
≥ ∂<br />
0<br />
T t<br />
2<br />
Check if<br />
threshold is crossed, then<br />
provide new period for damping pr<strong>of</strong>ile ( T = T j<br />
).<br />
0<br />
We take<br />
8 sections.<br />
∂t = T 0<br />
8<br />
since we divide the gait cycle into<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
Classification <strong>of</strong> eight<br />
regions over a gait cycle,<br />
superposed over the actual<br />
data <strong>of</strong> the knee angle.
RF Module<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
RF Module<br />
• An RF modulator<br />
(for radio<br />
frequency<br />
modulator) is a<br />
device that takes a<br />
baseband input<br />
• signal and outputs<br />
a radio frequencymodulated<br />
signal.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Why did nature never invent<br />
the wheel<br />
• Well, Nature has exploited circular objects and<br />
surfaces in many <strong>of</strong> its designs. Many<br />
mammalian joints follow the classic ball-andsocket<br />
design that allows circumduction (full 360<br />
degree range <strong>of</strong> movement).<br />
• When we try to mimic nature<br />
(Legged motion) we get in to a lot<br />
<strong>of</strong> problems related to stability and<br />
energy efficiency.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Legged locomotion<br />
• Adaptability and maneuvrability in rough<br />
terrain.<br />
• Mechanical complexity.<br />
• Less efficient in energy.<br />
• Complexity in control.<br />
• Nature has usually preferred legs.<br />
• Stability is a problem.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong> - <strong>Allahabad</strong>
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Monoped locomotion<br />
• Makes continuous<br />
adjustments to leg<br />
angle to control<br />
the body attitude<br />
and velocity.<br />
• Dynamic balance<br />
only.<br />
(Stability)<br />
http://www.ai.mit.edu/projects/leglab/<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Biped locomotion<br />
• Static balance is available in a small range.<br />
Dynamic stability required for movement.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Outline<br />
Objective<br />
Contribution up to now<br />
Design <strong>of</strong> IGOD<br />
Analysis <strong>of</strong> human gait oscillations<br />
Application<br />
Conclusion and Future Work<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
9
The dotted line is a simple sine curve fit to the shoulder elbow<br />
and hip oscillations.<br />
θ<br />
L<br />
= θ<br />
L<br />
sin( ωt + δ<br />
L<br />
)<br />
i 0i<br />
i<br />
R<br />
R<br />
0i<br />
θ = sin( +<br />
θ ωt δ<br />
i<br />
As expected, we notice that the periods <strong>of</strong> oscillation for the fitted joints are<br />
the same ( T = 2 π ω =1. 57 sec). It is essential for the stability that all limbs<br />
should oscillate at the same frequency. We also notice as expected that the<br />
phase difference between the corresponding left and right limb is<br />
∆δ = δ −δ<br />
=<br />
i<br />
L<br />
i<br />
for the shoulder and hip joints. To check the correlation and coupling<br />
between the left and right <strong>of</strong> the body during a stable gait, and the<br />
consistency <strong>of</strong> the data, we plot the left with the right limb oscillations<br />
L<br />
i<br />
π<br />
R<br />
i<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
)
Degrees <strong>of</strong> Freedom (DOF)<br />
The arrows represent<br />
the unidirectional<br />
coupling <strong>of</strong> this first<br />
model. The values on<br />
the arrows correspond<br />
to the phase<br />
differences we<br />
imposed between the<br />
different oscillators.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Figure 7. Correlations and coupling between the left and right<br />
limbs. In this figure, the fitted sine functions (in Fig. 6),<br />
describe a Lissajous curve, shown in thick lines.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Contribution up to now<br />
• Rotation sensor based biometric wearable suit has been made<br />
for capturing different human gait oscillations for analysis<br />
and different applications.<br />
• Introduce rotation sensor base biometric wearable suit for<br />
human identification.<br />
• Personal recognition system which can applicable to real<br />
world situation.<br />
• Better feature selection procedure which can help us to<br />
recognize the person with giving 100% accuracy.<br />
• Choose the real time pattern<br />
compatible with our feature.<br />
recognition algorithm<br />
• Better accuracy<br />
approaches.<br />
has been achieved rather existing<br />
• Introduce Tree Structured LDA (TS-LDA) classification<br />
algorithm.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Application<br />
1. Gait Based Personal Recognition<br />
System<br />
2. Biological controller<br />
development for biped robot to<br />
generate biped locomotion<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
Gait Based Personal Recognition<br />
System<br />
To recognize different human walking in real time<br />
for Identification and research.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
Recognition Procedure<br />
Person<br />
Movement<br />
Rotation sensors<br />
based suit<br />
( IGOD)<br />
Capture Joint<br />
Angle values<br />
Create Feature<br />
Matrix<br />
Choose 60%<br />
sample for<br />
training<br />
Recognition<br />
using ANN/TS-<br />
LDA<br />
After training rest<br />
<strong>of</strong> data taken for<br />
testing<br />
Result<br />
106<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Data collection and<br />
Feature Extraction<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
Data collection<br />
1. We have taken 30 (25 male and 5 female) different humans 30sec<br />
normal walking data and the data was taken in 10 different sessions.<br />
2. The age <strong>of</strong> our volunteers is 20-30 years.<br />
3. According to our different volunteers point <strong>of</strong> view the suit does not<br />
affect their natural walking.<br />
4. In each session we have taken just 30 sec <strong>of</strong> walking step in order to<br />
<strong>of</strong>fer some amount <strong>of</strong> freedom to their normal walking. It leads to an<br />
assumption that if we put any constraint such as only 4 gait cycle<br />
needs to be accounted while capturing walking pattern, it will affect<br />
tremendously from his/her natural gait pattern.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
Phases <strong>of</strong> the gait cycle<br />
Stance Phase<br />
• Initial Double Support<br />
• Single Support<br />
• Terminal Double Support<br />
Swing Phase<br />
• Initial Swing<br />
• Terminal Swing<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
Feature vector<br />
• Data Folding is used for creating feature<br />
vector.<br />
• Minimum and Maximum angle value <strong>of</strong><br />
each joints within a particular Gait Cycle.<br />
• Time taken by the Gait Cycle.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
Feature vector…<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
Feature vector…<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
Gait Recognition<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
Method Used<br />
• Artificial Neural Network (ANN)<br />
• Tree Structured Linear Discriminant<br />
Analysis (TS-LDA)<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
LDA<br />
• For detecting the variables that allow the<br />
researcher to discriminate between different<br />
(naturally occurring) groups.<br />
• For classifying cases into different groups with a<br />
better accuracy.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
LDA Numerical Example<br />
Factory “ABC” produces very expensive and high quality chip rings that heir qualities are<br />
measured in term <strong>of</strong> curvature and diameter. Result <strong>of</strong> quality control by experts is given in<br />
the table below.<br />
Curvature Diameter Quality Control Result<br />
2.95 6.63 Passed<br />
2.53 7.79 Passed<br />
3.57 5.65 Passed<br />
3.16 5.47 Passed<br />
2.58 4.46 Not Passed<br />
2.16 6.22 Not Passed<br />
3.27 3.52 Not Passed<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
Solution<br />
X= features (or independent variables) <strong>of</strong> all data. Each row (denoted by) represents<br />
one object; each column stands for one feature.<br />
Y= group <strong>of</strong> the object (or dependent variable) <strong>of</strong> all data. Each row represents one<br />
object and it has only one column.<br />
In our example, and<br />
2.95 6.63<br />
2.53 7.79<br />
X= 3.57 5.65 Y=<br />
3.16 5.47<br />
2.58 4.46<br />
2.16 6.22<br />
3.27 3.52<br />
1<br />
1<br />
1<br />
1<br />
2<br />
2<br />
2<br />
1<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Solution…<br />
X i = features data for group i. Each row represents one object; each<br />
column stands for one feature. We separate X into several groups<br />
based on the number <strong>of</strong> category in Y.<br />
2.95 6.63<br />
2.53 7.79<br />
X 1 = X 2 =<br />
3.57 5.65<br />
3.16 5.47<br />
2.58 4.46<br />
2.16 6.22<br />
3.27 3.52<br />
μ i = mean <strong>of</strong> features in group i, which is average <strong>of</strong> X i<br />
μ 1 = μ 2 =<br />
3.05 6.38 2.67 4.73<br />
μ = global mean vector, that is mean <strong>of</strong> the whole data set.<br />
μ =<br />
2.88 5.676<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
Solution…<br />
X’ i = mean corrected data, that is the features data for group<br />
i => X i , minus the global mean vector μ .<br />
0.06 0.951<br />
- 0.357 2.109<br />
X’ 1 = X’ 2 =<br />
0.679 - 0.025<br />
0.269 - 0.209<br />
- 0.305 - 1.218<br />
- 0.732 0.547<br />
0.386 - 2.155<br />
C i = [(X’ i ) T * X’ i ] / n i = covariance matrix <strong>of</strong> group i.<br />
C 1 = C 2 =<br />
0.166 - 0.192<br />
- 0.192 1.349<br />
0.259 - 0.286<br />
-0.286 2.142<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
Solution…<br />
= pooled within group<br />
covariance matrix. It is calculated for each entry (r, s) in the matrix. In<br />
our example,<br />
C =<br />
0.206 - 0.233<br />
- 0.233 1.689<br />
The inverse <strong>of</strong> the pooled covariance matrix is<br />
C -1 =<br />
5.745 0.791<br />
0.791 0.701<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
Solution…<br />
P i = prior probability vector (each row represent prior probability <strong>of</strong><br />
group i ). If we do not know the prior probability, we just assume it<br />
is equal to total sample <strong>of</strong> each group divided by the total samples,<br />
that is P i = n i / N<br />
P =<br />
4 / 7<br />
3 / 7<br />
Discriminant function<br />
f i = μ i C -1 X k<br />
T<br />
– (μ i C -1 μ i<br />
T<br />
) / 2 +ln ( P i )<br />
We should assign object k to group i that has maximum f i<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
Solution…<br />
The discriminant function is our classification rules to assign the object into<br />
separate group. If we input the new chip rings that have curvature 2.81 and<br />
diameter 5.46, reveal that it does not pass the quality control.<br />
Transforming all data into discriminant function (f 1 , f 2 ) we can draw<br />
the training data and the prediction data into new coordinate.<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
Recognition Procedure with TS-<br />
LDA<br />
X<br />
R 1 R 2<br />
Max (f 1 , f 2 ) = f 1<br />
R 3 R 4<br />
Max (f 3 , f 4 ) = f 4<br />
X<br />
R 1 R 4<br />
X<br />
Max (f 1 , f 4 ) = f 4<br />
X is recognized<br />
as class 4<br />
123<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong>
Result analysis<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1
Performance Evaluation<br />
No. <strong>of</strong> Subjects<br />
Performance<br />
Gafurov et al. [3][10] 21 5%-13% EER<br />
Nakajima et al. [4] 10 85% RR<br />
Middleton et al. [5] 15 80% RR<br />
Derawi et al. [7], [8] 60 5.7%-20% EER<br />
Rong et al. [9] 35 6.7% EER<br />
Pan et al. [18] 30 96.7% RR<br />
Moustakidis et al. [19] 40 72.89% - 98.21% RR<br />
Our approach 30 100% RR<br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong>, <strong>Allahabad</strong><br />
<strong>Indian</strong> <strong>Institute</strong> <strong>of</strong> <strong>Information</strong> <strong>Technology</strong> - <strong>Allahabad</strong><br />
1