1. PIR Based Non-Cooperative Localization
Wireless Sensor Network
Major Project-II Report
Submitted in partial fulfillment of the requirements for the award of the degree
Bachelor of Technology
In
Electrical and Electronics Engineering
By
Bhavana M.C. (09EE27),
Nikhil S. (09EE61),
Siddhartha Kumar (09EE91),
Vivekananda S. (09EE104)
Under the guidance of
Dr. Ashvini Chaturvedi,
Associate Professor,
Department of Electrical and Electronics Engineering,
2. National Institute of Technology Karnataka, Surathkal
SRINIVASNAGAR-575025, KARNATAKA, INDIA
DECEMBER, 2012
Contents
Topic Pg.no
1. Declaration..……………………………………………………………………………………………………….3
2. Certification………………………………………………………………………………………………………..4
3. Acknowledgement ……………………………………………………………………………………………..5
4. Abstract…………………………………………………………………………………………………………..…6
5. Introduction……………………………………………………………………………………………………....7
6. Literature survey…………………………………………………………………………………………………8
7. Problem statement……………………….……………………………………………….…………………..10
8. Proposed methodology of solution………………………………………………….………………….10
9. Expected outcomes………..………………………………………………………………..…………………10
10. Pyro Electrics Infrared Sensors and Analog circuit design…………..…….………….…….11
11. SVM classification………………..…………………………………………………………….……………..20
12. WSN localization using PIR………………………………………………………………….…………….28
13. Conclusion…………………………………………………………….…………………………….……….…..29
14. References……………………………………………………………..………………………….………….….30
15. Appendix A…………………………………………………………………………………………………………31
16. Annexure 1………………………………………………………………………………………………………...32
17. Annexure 2…………………………………………………………………………………………………………33
3. DECLARATION
We hereby declare that the project work report entitled “PIR Based Non-Cooperative Localization
Wireless Sensor Network” which is being submitted to the National Institute of Technology
Karnataka, Surathkal for the award of the Degree of Bachelor of Technology in Electrical and Electronics
Engineering is a bonafide report of the work carried out by us. The material contained in this report has
not been submitted to any university or institution for the award of any degree.
SI. NO. NAME ROLL NO. Signature
1 Bhavana M.C. 09EE27
2 Nikhil S. 09EE61
3 Siddhartha Kumar 09EE91
4 Vivekananda S. 09EE104
Department of Electrical and Electronics Engineering
PLACE: NITK, SURATHKAL
DATE: 6th December, 2012
4. CERTIFICATE
This is to certify that the B-Tech project work report entitled “PIR Based Non-Cooperative
Localization Wireless Sensor Network” submitted by:
SI. NO. NAME ROLL NO.
1 Bhavana M.C. 09EE27
2 Nikhil S. 09EE61
3 Siddhartha Kumar 09EE91
4 Vivekananda S. 09EE104
As the record of the work carried out by them, is accepted as the B-Tech. project work report
submission, in partial fulfilment of the requirements for the award of degree of Bachelor of Technology
in Electrical and Electronics Engineering.
Dr. ASHVINI CHATURVEDI
Project Guide,
Department of Electrical and Electronics Engineering,
National Institute of Technology Karnataka, Surathkal
Dr. K.P VITTAL
Head of the Department,
Department of Electrical and Electronics Engineering,
National Institute of Technology Karnataka, Surathkal
5. ACKNOWLEDGEMENT
The extensive endeavour, bliss and euphoria to accomplish the task certainly would not have
been realized without the expression and gratitude of people who made it possible. We take this
opportunity to acknowledge all those whose support and encouragement has helped us in tuning
this project.
We are grateful to our guide, Dr. Ashvini Chaturvedi, department of Electrical and Electronics
Engineering for not only providing us opportunity to showcase but also all facilities and
experience in the completion of this project. He bestowed his guidance at appropriate times
without which it would have been very difficult for us to complete the project. An assemblage of
this nature could never have been attempted without the support of our guide.
We would like to thank Prof. N.S.V. Shet and Centre for Excellence- Wireless Sensor Network
for extending their support in completion of our project.
Our thanks are also due to Prof. K.P. Vittal, the head of the Electrical and Electronics department
who has allowed us use of the facilities at the department.
Bhavana M.C. (09EE27)
Nikhil S. (09EE61)
Siddhartha Kumar (09EE91)
Vivekananda (09EE104)
6. Abstract
This project is an implementation of a wireless sensor network to demonstrate localization using
pyro-electric infrared sensors. Our project is divided into two main parts; the use of analog
output from the PIR node to determine accurate location of an individual and use of digital
output from a comparator circuit for low computation approximate localization. The analog
output from the PIR circuit is run through a feature extraction algorithm and then the feature
points are sent to a neural classifier for distance classification. This data can be used to classify
distance or speed of the individual. The digital output can only determine the presence or
absence of the individual in the field of view of that sensor. The wireless sensor network uses a
multi-hop Zigbee network to transmit the data.
7. Introduction
Pyroelectric Infrared (PIR) sensors belong to the class of thermal detectors. Thermal detectors
can measure incident radiation by means of a change in their temperature. When an appropriate
absorbing material is applied to the detectors element surface, they can be made responsive over
a selected range of wavelengths. PIR sensors are designed to detect human bodies, thus the
wavelengths of interest are mainly in the range of the IR window at, in which the IR emission of
bodies at 37 also peaks. Being low-cost, low-power, and providing a reliable indication of people
presence, PIR sensors have achieved worldwide diffusion. Furthermore, they can be
manufactured with a reduced form factor that allows to unobtrusively integrating a large number
of them around us. Nowadays, many buildings include automatic light switching and
surveillance system based on a large number of PIR scattered in different rooms. Beyond simple
presence, the output of a PIR sensor depends on several characteristics of the body moving in its
field of view, such as direction of movement and distance of the body from the sensor. This
observation has motivated our effort in developing a novel technique to extract these features. In
particular, our objective is to implement a human tracking system based on a dense array of PIR
sensors. Previous works demonstrated how such system can be used to improve video
surveillance systems and preserve privacy.
In this paper, we present a technique to track people using an array of PIR sensors distributed in
the environment. This technique requires low computational power, is suitable for a parallel
implementation, and is based only on low-cost, low-power devices. Hence, it is well suited for
implementation on wireless sensor network (WSN) nodes, further reducing the obtrusiveness and
cost since no wires are needed.
8. Literature Survey
Infrared (IR) radiation is a type of electromagnetic radiation. Infrared light has a longer
wavelength than visible light. The infrared has a wavelength of 750 nm to 100 μm. The infrared
radiation is invisible to humans but we can feel it as heat. According to the thesis ‘Pyroelectric
Infrared (PIR) Sensor Based Event Detection’ by Emin Birey Soyer[2], Pyroelectricity is the
ability of certain materials to generate a temporary electrical potential when they are heated or
cooled. These changes in the heat are produced when a warm/hot bodied object passes by those
materials. Hence PIRs can be used for motion detection. PIR sensor is an electronic device that
generates an electric charge when exposed to infrared radiation. As the name implies this sensor
is made of pyroelectric materials such as crystals. When the amount of infrared radiation that is
striking to the crystal changes, the amount of charge also changes. This charge is sensed and
converted to a voltage level via a FET transistor that is build inside the sensor. The sensor is
sensitive over a wide spectrum. A Fresnel lens attached to the PIR improves the field of vision
for detection. A Fresnel lens is a plano convex lens that has been collapsed on itself to form a at
lens that retains its optical characteristics but is much smaller in thickness and therefore has less
absorption loss. The Fresnel lens is made of an infrared transmitting material that has an IR
transmission range of 8 μm to 14 μm that is most sensitive to human body radiation. Another
issue of a Fresnel lens is the pattern. The pattern affects the performance of the sensor directly.
In order to increase the performance and coverage, different geometries are applied.
We first studied the result on analogue output of the PIR when a warm object was waved in front
of the PIR. An oscilloscope was used in the mentioned process, and we could detect faint
impulses generated by the sensor. The output was in terms of millivolts and thus the choice of an
oscilloscope instead of a digital multimeter. Also we could observe the shape of the response
generated, vaguely. We checked the PIR output with and without the Fresnel lens, and the range
was a little markedly increased with its use
Since PIR sensor's output is very low, it should be well amplified. For processing and utilization
of a PIR output in a microcontroller, amplification is necessary. A suggested gain of around
10000 would suffice microcontroller’s utilization. Because there is a high gain, amplifiers with
band-pass characteristics are used. The output of a PIR sensor has a low frequency tendency
around DC to 10Hz. So the frequency response of the amplifier block tends to remove or reduce
the high frequency components while amplifying the interested frequency band in the output
signal of a PIR.
We used active band pass filters for amplification and band limiting. A cascaded filter circuitry
was essential in getting a high gain. A good amplification implies that the microcontroller can
effectively use its Analogue-Digital conversion for signal processing and a good band limitation
implies increased sensitivity. This way ambient noise could be eliminated. We tried two different
amplifications and two different bandwidths. The circuit with more amplification, of the two and
lesser bandwidth was found to be more viable.
9. PIRs can be used for three purposes:
Direction of movement: In the presence of a single lens, the passage of a body results in a PIR
output signal made up of two peaks, one positive and one negative. The reason is that the sensing
elements detect the body in sequence. Being placed in series with opposite polarization each of
them causes a peak with different direction.
Distance of movement: The signal amplitude (calculated as the difference between the
maximum and minimum value of the PIR output) is at a maximum for passages in the smaller
distance.
Speed of movement: The signal duration (calculated as the time between the instant when the
PIR output exceeds one of the two thresholds and the instant when it lies between the thresholds
for some time) increases with decrease in speed.
We took many trials for various distances and different speeds. The acquired data was almost
analogous to the features explained in the paper. A tabulation of all the collected data was made
for further classification process.
Classification of distance and speed is possible when the above mentioned voltage difference and
time difference can used as suitable parameters. Below mentioned are three suggested types of
classification.
1) Naïve Bayes: The Naïve Bayes classifier is a simple probabilistic classifier based on Bayes
theorem, and the assumption that input features are independent. Using the Bayes theorem, the
classifier calculates the posterior probability of all classes given the input features. A decision
rule selects the output class: in this paper, we assign the instance to the class with higher
posterior probability.
2) -Nearest Neighbour (k-NN): -NN, given a set of reference instances, classifies a new pattern
as the one most represented among the closer ones [26]. -NN training phase is simply the
collection of a set of reference instances from each class. The drawback of this approach is that
its complexity and memory cost increase with reference dataset dimensions, which may be
relatively large. Moreover, the accuracy of the algorithm can be severely limited by noisy
training instances, especially if k is small (i.e., k=1).
3) Support Vector Machines: SVM belongs to the class of linear discriminant classifiers. Such
classifiers use discriminant functions that are a combination (either linear or not linear) of the
input vectors’ components. Geometrically, a discriminant function defines a hyperplane that
separates two classes. Several solutions have been proposed to deal also with non-separable data.
The SVM use a set of kernel functions to pre-process the input vectors and represent them in a
higher dimensional space where they can be separated more easily. The training phase looks for
the support vectors, which are the (transformed) training instances closer to the separating
hyperplanes and are used to build the hyperplanes for the classification.
We tried k-NN and Naïve Bayes classification initially in the intent of performing dynamic
learning. The separate domains after training were not very well demarcated as it is a
probabilistic process. Hence SVM was used, owing to its linearity and simple classification data
for fresh data points. Here different domains are separated by straight lines, which make
classification lucid enough for a microcontroller
10. Problem Statement
In today's world; there is an utmost demand for effective use of surveillance equipment in
densely crowded areas, which are insecure and almost impossible to monitor by security
personnel alone. These areas are usually public places namely banks, railway stations, bus
terminals, libraries, exhibitions etc., and are characterized by exceptionally high population
density during specific time windows. The surveillance of such areas is not an easy task and has
to be designed for non-cooperative situations as well.
Proposed Methodology of Solution
Our approach for the aforementioned problem of surveillance is to use cheap Pyroelectric
Infrared (PIR) sensors as a wireless sensor network mesh throughout the area. The PIR sensors
are capable of detecting and tracking motion. These PIR sensor arrays will act as a simple trigger
circuit that can constantly monitor the motion of the crowd and in turn feeds the system with
information of where most people are gathered. Added features of trying to identify any
abnormal movement or objects can also be implemented easily. On gathering preliminary
information via the PIR wireless sensor network net, we can direct the overhead surveillance
camera towards the area of the crowd and run a crowd flow analysis. Image processing to
determine the nature of the gathering and detect any abnormalities if required. By narrowing the
area of surveillance we are allowed to optimize the utility and focus our analysis to the restricted
area
Expected Outcome
Complete prototype of filter circuit for the PIR sensor array
Design PCB for PIR sensor circuit and Test sensor module
Build sensor module microcontroller and Zigbee transceiver
Use Support vector machine classifier to obtain boundary equations for speed and
distance
Verify Results with test cases.
Modify filter circuit for digital output
Program microcontroller to take in digital output of motion detection and relay the data to
the Zigbee network
Obtain a Visual representation of the tracking data from multiple nodes.
11. Pyro-electric Infrared Sensors and Analog Circuit
Design
Pyroelectricity is the electrical response of a polar, dielectric material to a change in its temperature. A
pyroelectric element converts incident IR flux into an electrical output through two steps: the absorbing
layer transforms the radiation flux change into a change in temperature and the pyroelectric element
performs a thermal to electrical conversion. Commercial-off-the-shelf (COTS) PIR sensors include two
sensitive elements placed in series with opposite polarization. This configuration makes the sensor
immune to slow changes in background temperature and reduces the settling- down period once input
radiation changes settle down.
Working of a Typical PIR sensor:
Strictly speaking, individual PIR sensors do not detect motion; rather, they detect abrupt changes in
temperature at a given point. As an object, such as a human, passes in front of the background, such as a
wall, the temperature at that point will rise from room temperature to body temperature, and then back
again. This quick change triggers the detection. Moving objects of identical temperature, however, will
not trigger detection because that is the sensors job. It detects motion regardless of temperature. This is
also why you can take a piece of paper from a desk that is the same, exact, temperature as everything in
the room, attach it to a stick (also the same exact temperature as the room) and wave it slowly in front of a
sensor (at any range within the limit of the sensor) and it will activate the sensor).
The PIR sensors are used in conjunction with Fresnel lenses to augment and shape their Field of vision
Fresnel lenses are good energy collectors that can be moulded out of inexpensive plastic and present a
much more compact form factor with respect to normal lenses. Typically, an array of Fresnel lenses is
used to divide the PIR sensor Field of Vision into several, optically separated cones. The motivation is
that the PIR elements detect only changes to incident IR radiation. If a single lens is used, as a body
moves through the Field of Vision of the PIR (especially if it covers a wide area) only, negligible changes
in input IR radiation will be sensed. On the other end, when using multiple lenses, the body moves
between different cones of view and is sensed for the whole traversal.
12. Figure: Schematic of a typical of the shelf Pyroelectric Infrared. Two sensing elements are used in series
with opposite polarization; the output is amplified through a double stage amplified band-pass filter. The
field of view of each sensing element is highlighted with shading. It is worth noting how, in proximity of
the device, the two Field of Vision overlap.
Specifications of PIR:
DC Output Voltage: 3.3mVp-p
Power Dissipation: 500mW
Operating temperature: -40°C to +125°C
Supply Voltage to ground potential (Vcc): -0.3V to 6V
Sensing angle: 41°
Sensing angle with Fresnel lens: 90°
Whenever a hot body moves in the field of vision of a PIR sensor, one IR sensing element of the sensor
gets triggered and it gives a positive/negative pulse and when the body moves further to the field of vision
of the next IR sensing element it gives an opposite pulse. Therefore this forms the output of a PIR sensor.
PIR Output plots for different direction:
0 100 200 300 400 500 600 700 800 900 1000
0
100
200
300
400
500
600
700
800
900
1000
13. BASIC SENSOR SIGNAL CHAIN
In all of the intelligent sensing applications discussed, we use a common sensor signal path, which
consists of the sensing element, signal conditioning circuitry to map the sensing element’s output signal
into a range that the remaining electronic circuitry can process, and then filtering circuitry to reduce or
eliminate electronic noise. This signal conditioning circuitry conditions the signals coming from the
sensor, and since a majority of sensors generate analog output signals, this circuitry is essentially analog
rather than digital.
0 100 200 300 400 500 600 700 800 900 1000
0
100
200
300
400
500
600
700
800
900
1000
14. SIGNAL CONDITIONING CIRCUIT
The specific signal conditioning circuits that are needed in a sensor application depend on the type of
sensor employed.
First, the signals generated by the sensor must be as free of noise as possible. Moreover, the frequency
content of the signal, in other words its bandwidth, must be limited to a certain range, based on some
constraints about which we will learn shortly. This often makes it necessary to use a Bandpass Filter.
Secondly, the signals generated by PIR sensor sensor have weak amplitudes. In order to process the signal
accurately, and also to make the system more robust to the effect of noise, the signal needs to be
amplified.
In addition to filtering and amplification, the need to convert the signal into digital form using an Analog-
to-Digital Converter, or ADC, adds some more signal conditioning needs. Besides amplifying the signal,
the signal might also need to be translated to suit different ADC voltage references. Also, many ADCs,
especially those contained inside an MCU or DSC, only operate on unipolar inputs; that is, the input
voltage cannot alternate between positive and negative levels with respect to the Ground. In such cases, a
Level Shifter is required.
THEORY ABOUT ACTIVE BANDPASS FILTER:
The principal characteristic of a Band Pass Filter or any filter for that matter is its ability to pass
frequencies relatively unattenuated over a specified band or spread of frequencies called the "Pass Band".
For a low pass filter this pass band starts from 0Hz or DC and continues up to the specified cut-off
frequency point at -3dB down from the maximum pass band gain. Equally, for a high pass filter the pass
band starts from this -3dB cut-off frequency and continues up to infinity or the maximum open loop gain
for an active filter.
However, the Active Band Pass Filter is slightly different in that, it is a frequency selective filter circuit
used in electronic systems to separate a signal at one particular frequency, or a range of signals that lay
within a certain "band" of frequencies from signals at all other frequencies. This band or range of
frequencies is set between two cut-off or corner frequency points labelled the "lower frequency" ( ƒL )
and the "higher frequency" ( ƒH ) while attenuating any signals outside of these two points.
Simple Active Band Pass Filter can be easily made by cascading together a single Low Pass Filter with a
single High Pass Filter.
16. In the circuit above, pin 1 and 3 of the PIR sensor are connected as referred from the data sheet to a
resister and capacitor pair and to ground respectively. A pull down resister of 100k is connected between
pin 2 and 3 or PIR.
2-Stage Filter/Amplification:
The output signal from PIR is directed to 2-stage bandpass filter. The first filter stage is in non inverting
mode for which the bandwidth is 0.53Hz, lower cutoff frequency being 1.06Hz and upper cutoff
frequency being 1.59Hz. The gain of this filter is 66.667. This filter stage is cascaded to another bandpass
filter of same bandwidth and gain ut with a little modification. This filter is connected in inverting mode
and a DC offset of 2.5 volts is added to the output but keeping the non inverting pin of the OpAmp at
2.5V. The main purpose of this 2-stage filter circuit was to get a higher amplification factor as the PIR
output is of very weak magnitude. The total gain is gain1 multiplied by gain2 which comes to 4444.44.
The gain in dB is 72.956.
Window Comparator:
The output from the second filter is connected to a window comparator. The reference voltage for these
comparators is set from two potentiometers. The comparators give a positive pulse when the output of
filter is below or above the reference voltage Vf2 and Vf1. The output from the comparator is used to
trigger the microcontroller whenever hot body moves across the sensor circuit.
PSpice SIMULATION OF FILTER CIRCUIT:
The circuit was simulated on PSpice before putting it on PCB. The design schematic is,
17. MODIFICATIONS MADE IN THE CIRCUIT:
Few modifications were necessary to match the requirements. The original circuit had a bandwidth of 2Hz
and gain was 45dB. The desired bandwidth is 0.5Hz and gain is 70dB. So the resister-capacitor
combinations of filter circuit were modified. The changes made were:
Resister R3 and R4 changed from 10k to 15k.
Capacitor C1 and C4 changed from 1000nF to 100nF.
THEORETICAL RESULT:
The new value of gain per band-pass filter as obtained is 66.667
Therefore the total gain after two stage amplification is, Av = 4444.44
Overall gain in Decibels is 72.956 dB
Band cut-off frequencies are 1.06Hz and 1.59Hz
PSPICE SIMULATION OUTPUT:
Total gain obtained in Decibels is 65dB
Roll-Off value of the output signal is -38dB/decade
Bandwidth approximately equal to 0.8Hz
With the changes made we could obtain the desired value of gain and bandwidth. The rolloff value of the
2-stage filter is -38dB/decade because, each filter stage is a first order bandpass filter and each
18. corresponds to a -20dB/decade. When 2 filters are cascaded their gains get multiplied and on logarithmic
scale they sum up. Therefore -20 and -20 sums up to give -40dB/decade.
THE BODE PLOTS OF THE FILTER CIRCUIT:
PLOT OF ORIGINAL FILTER CIRCUIT:
BANDWIDTH OF ORIGINAL CIRCUIT:
20. SVM Classification of PIR feature points
Introduction:
Support Vector Machines – These are supervised learning models with associated learning algorithms that
analyse data and recognize patterns, used for classification and regression analysis. a support vector
machine constructs a hyperplane or set of hyperplanes in a high- or infinite-dimensional space, which can
be used for classification.
Tabulations of the acquired raw data:
Peak to
Peak
Voltage
Difference
Time
Difference
corresponding
to the Voltage
Difference
Distance
Group
Speed
Group
Peak to
Peak
Voltage
Difference
Time
Difference
corresponding
to the Voltage
Difference
Distance
Group
Speed
Group
606.8 153 1 1 545.4 43 1 1
305.2 91 1 1 324.6 95 1 1
373.4 50 1 2 236.8 30 1 2
257.6 75 1 2 368.8 83 1 2
330.6 30 1 3 247.6 36 1 3
310.4 32 1 3 387.2 66 1 3
320.2 51 2 1 282.2 91 2 1
255 65 2 1 325 80 2 1
224.4 81 2 2 302.6 39 2 2
244.8 37 2 2 290.6 55 2 2
171.4 56 2 3 209.2 76 2 3
182.8 52 2 3 203.6 31 2 3
105.4 123 3 1 101 119 3 1
170.8 122 3 1 121.8 65 3 2
153.2 70 3 2 107.4 134 3 1
162 69 3 2 112 100 3 2
122 53 3 3 142 60 3 3
123.2 43 3 3 130.2 57 3 3
85.6 93 4 2 59 135 4 1
114.4 68 4 2 55.6 186 4 1
75.2 122 4 1 66.4 116 4 2
98 90 4 1 63.2 110 4 2
106.2 64 4 3 79.4 79 4 3
128.8 57 4 3 72.2 52 4 3
573.6 96 1 1 664.2 80 1 1
363.6 61 1 1 355.8 107 1 1
21. 452.8 127 1 2 250.6 54 1 2
242 77 1 2 620.2 74 1 2
339.4 89 1 3 286.4 27 1 3
241 47 1 3 585 33 1 3
304.4 75 2 1 302.2 60 2 1
291 89 2 1 364.6 72 2 1
262 50 2 2 237.2 42 2 2
231.4 32 2 2 279.6 58 2 2
167.4 77 2 3 243.8 34 2 3
161 88 2 3 322.2 33 2 3
90.4 113 3 1 152.6 113 3 1
102.8 133 3 1 128.2 123 3 1
125 78 3 2 180.4 54 3 2
144 62 3 2 142 74 3 2
121.6 46 3 3 183 45 3 3
134.2 51 3 3 163 53 3 3
52 105 4 1 95.2 136 4 1
72.2 120 4 1 75 90 4 1
67.6 90 4 2 82.2 126 4 2
95 70 4 2 89.8 94 4 2
80.2 64 4 3 105.6 66 4 3
98.6 56 4 3 113.2 63 4 3
Training using SVM in MATLAB:
The above shown raw data was segregated using two different criteria: Peak to Peak Voltage Difference
(Distance group) and Time Difference (Speed group)
All the points belonging to one particular distance group are put together. In the above case, each of
groups 1, 2, 3, 4 are segregated and contain certain number of data points each.
Similar process is done for speed groups also. In the above case, each of groups 1, 2, 3 are segregated.
SVM training is binary, i.e. only two groups can be trained at once. Hence pairwise training of the above
groups is done.
Speed groups 1 & 2 are done first, then 2 & 3, and followed by 3 & 4. For time groups, the classification
of group 2 is not very crisp, hence only groups 1 & 3 are trained for better results.
A linear classification is used in this case for simplicity. MATLAB does not give the equation of the
separating line (between two groups/classes). Hence the equation of this particular line has to be found
out manually. The function ‘ginput’ has been used in MATLAB to extract two points in a line, which
suffice the requirements to find the equation of that line. A standard form of ‘ax + by + c = 0’ has been
incorporated to find the line equation’s parameters a, b and c.
The equations of these lines are quintessential in classification of fresh data. We can find out the side of
the line in which any particular point on a graph lies, using the same equation ax + by + c = 0, with
22. substitution of x & y of that point. Locating the point between two particular separation lines helps us find
the group to which a fresh data point lies.
Plots:
The results obtained from SVM training can be graphically plotted.
SVM training for distance classes 1 & 2:
23. SVM training for classes 2 & 3:
SVM training for distance classes 3 & 4:
24. Combination of the different distance groups:
SVM training for speed classes 1 & 3:
28. Wireless Sensor Network Localization Using PIR
Previously we have analyzed how we can use the conditioned analog output from our PIR sensor module
to be able to classify the persons speed and distance and hence obtain a higher level of understanding the
crowds or intruders activities in the area under surveillance. But the analog output consists of about 1
million samples for a second (1Mhz sampling frequency for 1 sec), each sample is of 10 bits. This much
amount data (10 million bits approximately) is too much to be sent over a wireless link with almost zero
delay. That’s why a compromise is made for almost real time detection of motion. For going the data of
speed and distance a simple approximation of the monitored individual being present or not in the field of
view of the PIR is what is obtained a simple 1 bit data from each sensor.
Implementation:
This concept is implemented by simply using a comparator circuit after the analog output the circuit used
is given in the below figure.
Figure. Digital Output circuit
Detection Circuit:
The Potentiometers adjust the reference voltage to the comparators, hence a positive peak for is detected
by U2 and the negative peak by U3. These two digital outputs are fed to the MSP430 interrupt pins.
29. Figure: Comparator Levels and output
Sensor Node:
The Sensor node (shown below) consists of a MSP430 microcontroller. The microcontroller waits for an
interrupt on a GPIO pin, connected to the PIR sensor module. Once this interrupt is detected immediately
a packet with the Node number and Pin number is sent via the Zigbee module to the Zigbee coordinator
Figure: Zigbee PIR Node
Network:
The Zigbee module used is from TI’s CC2530 Mini-ZNP kit. The Zigbee network is setup by the Zigbee
coordinator on the Beaglebone. Once the network is setup each node when powered on connects to the
Zigbee coordinator (original parent) or finds a router (parent) to become a child of and eventually
connecting to the coordinator via a Multi-Hop Network. The Zigbee coordinator transmits the data
received from the Zigbee Sensor network, via Wi-Fi so that it can be accessed by any Wi-Fi capable
device that can connect to a static IP server. This is made possible as the ARM processor on the
Beaglebone runs angstrom that hosts a server where the data is sent to, any client that connects to this
server from different ports can access the sensor network data via Wi-Fi, hence expanding the capabilities
of the entire sensor network.
31. Matlab Software Algorithm:
Figure: Block Diagram of MATLAB Algorithm
The MATLAB algorithm takes into input values from the Zigbee network, Node number and Pin
Number. It plots the coverage area of the specific PIR depending on its known location in the area, hence
by highlighting multiple triggers one can localise an individual using intersecting regions.
Figure: MATLAB GUI a) Node 1 Field of view b) Node 2 Field of view
Results:
Once the Zigbee network was setup upon triggering of Node 1 we would see the plot update to show the
highlighted field of view that is monitored by that node, as shown in the above figure. Similarly on
triggering of Node 2 its respective field of view is highlighted. The grid represents an 8 feet X 8 feet grid
that was used for testing purposes; the nodes were placed in the middle opposite to each other.
32. Conclusion
Our project aims to use a simple passive sensor like the pyroelectric infrared sensor and
implement an entire system for tracking, locating and verifying movements of human beings. We
were able to start with a simple amplified and band passed analog signal extraction from the PIR.
We then verified the results of [1] , where the effects of distance and direction were observed on
the analog signal. We used K-nearest neighbor, Bayes and Support Vector Machine classifiers to
extract and classify the information of speed and distance of the individual. By using the SVM
linear boundaries were defined between each class and equations were used to simply verify
which class a new test case belongs to. This allowed us to reduce the memory requirement on the
sensor node side and hence allow us to greatly reduce the load on the network.
For real time and low cost sensor network we fed the analog signal to a digital output giving rise
to a 1 bit digital signal to indicate the presence or absence of motion of a human being. By using
multiple PIR’s in an array we can use this simple 1 bit data to track the movements of an
individual to certain extent, the more number of sensors the more accurate our results. For the
sensor network we used a Multi-Hop Zigbee network which then fed the data to a Wi-Fi server
on an ARM processor this allowed us to extend the range of the Sensor network from the base
station as well as allow multiple off the shelf devices like phones and laptops to interface with
the Zigbee network. The data obtained was then plotted on a MATLAB generated grid that in
practice would be overlaid over the area monitored, this would allow us to visualize the triggers
and detections of an individual in the area.
33. References
[1] Piero Zappi, Elisabetta Farella, and Luca Benini,” Tracking Motion Direction and Distance
With Pyroelectric IR Sensors”, IEEE Sensors Journal, VOL. 10, NO. 9, September 2010
[2]Emin Birey Soyer, “PIR sensor based event detection ” a thesis submitted to the Department of
Electrical and Electronics Engineering and the institute of engineering and sciences of Bilkent University
in partial fulfilment of the requirements for the Degree of Master of Science, July 2009
[3]R. Cucchiara, A. Prati, R. Vezzani, L. Benini, E. Farella, and P. Zappi, “Using a wireless sensor network
to enhance video surveillance,” J. Ubiquitous Comput. Intell. (JUCI), vol. 1, pp. 1–11, 2006.
[4] U. Gopinathan, D. Brady, and N. Pitsianis, “Coded apertures for efficient pyroelectric motion
tracking,” Opt. Exp., vol. 11, no. 18, pp. 2142–2152, 2003.
[5] Q. Hao, D. Brady, B. Guenther, J. Burchett, M. Shankar, and S. Feller, “Human tracking with wireless
distributed pyroelectric sensors,” IEEE Sensors J., vol. 6, no. 6, pp. 1683–1696, Dec. 2006.
[6] M. Shankar, J. B. Burchett, Q. Hao, B. D. Guenther, and D. J. Brady, “Human-tracking systems using
pyroelectric infrared detectors,” Opt. Eng., vol. 45, no. 10, pp. 106401-1–106401-10, 2006.
[7] N. Li and Q. Hao, “Multiple human tracking with wireless distributed pyro-electric sensors,” Proc. SPIE,
vol. 6940, no. 1, pp. 694033-1–694033-12, 2008.
35. Annexure 1
Laboratory Setup:
5 Sensor Nodes with 15 sensor modules
Component Manufacturer Cost per
component
Quantity Total cost of
component
1 LM324 Quad Opamp TI $0.45 15 $6.75
2 MSP430G2553 TI $2.25 5 $11.25
3 CC2530 Mini ZNP kit TI $99 2 $198
4 LM317KTTR TI $0.84 5 $4.20
5 Beaglebone Circuitco $80 1 $80
6 Wi-fi dongle Asus $18 1 $18
7 USB hub Croma $6 1 $6
8 PIR Sensors Murata $1.2 15 $18
9 PCB Fabrication Chipkraft $1.66 15 $25
10 Miscellaneous
discrete components
$20
Total cost of the project $387.2
Approximate Rs. 21,300
36. Annexure 2
Railway Platform Setup
Features:
1. Coverage Area: 200 feet Platforms covering 5 Platforms
2. Communication: Using VHF radio setup for the Speaker and Walkie-Talkie system
available at all Indian Railway stations
3. Control: Control Signals to Television sets, Speaker Systems and CCTV cameras, with
Pan and Tilt Capabilities.
4. Graphical User Interface Design: For easy visualization of data at control station.
Component (Package) Manufacturer Cost per
component
Quantity Total cost of
component
Comments
1 TLE2084A (SOIC) TI $3.00 400 $1200 JFET Fast response Op-amp
2 TM4C123GH6PM (LQFP) TI $4.50 200 $900 Microcontroller capable of SVM
3 TRF6901(LQFP) TI $4.50 200 $900 Single-Chip RF Transceiver
4 AC/DC PS 1PH404 Aegis PS $200 5 $1000 3.3V at 45 A Supply
5 Control Station - - - - Already Available
6 A3967 (SOIC) Allegro $0.50 100 $50 Control of Pan Tilt Motor
7 PIR Sensors Murata $1.2 400 $2800 High Sensitivity PIRs
8 PCB Fabrication - $1.00 400 $400 Designed for packages specified
9 Miscellaneous Costs for
GUI and Infrastructure
$2000 Includes: Wire laying, Secure
boxes for sensors, Networking
Total cost of the project $9250
Approximate Rs. 5,03,000