Coefficient of Thermal Expansion and their Importance.pptx
Intrusion Detection In Open Field Using Geophone (Report)
1. 1
INTRUSION DETECTION IN OPEN FIELD
USING GEOPHONE
A dissertation submitted in partial fulfillment for the requirement of
Awarding the degree of
Master of Technology
In
ROBOTICS ENGINEERING
Submitted By
Nuthan Prasad KB
(R860213009)
Under the Guidance of
Prof. Rajesh Singh
Assistant Professor, Department of Electronics and Instrumentation
Engineering
COLLEGE OF ENGINEERING STUDIES,
UNIVERSITY OF PETROLEUM AND ENERGY STUDIES,
DEHRADUN
MARCH 2015
2. 2
CERTIFICATE
This is to certify that the dissertation entitled “INTRUSION DETECTION IN OPEN
FIELD” is the bona fide record of the project work carried out by Mr. NUTHAN PRASAD
KB (R860213009) under my supervision and guidance in partial fulfilment of the
requirement for the award of the degree of MASTER OF TECHNOLOGY.
Further to the best of my knowledge, this report has not been submitted in parts or full to
any other University or Institute for the award of any other degree or diploma.
DATE: Prof. Rajesh Singh
Place: UPES, Dehradun
3. 3
DECLARATION
I, NUTHAN PRASAD KB, hereby declare that the project work which is being presented
in this dissertation titled “INTRUSION DETECTION IN OPEN FIELD” is an authentic
work carried out by me at University Of Petroleum And Energy Studies, Dehradun, under
the guidance of Prof. Rajesh Singh, College of Engineering Studies, UPES, Dehradun in
partial fulfilment of the requirements for the award of degree of Master of Technology in
Robotics Engineering. This report has not been submitted anywhere else for the award of
any other degree/diploma.
Date: Nuthan Prasad KB
Place: (R860213009)
4. 4
ACKNOWLEDGEMENTS
The completion of any project work depends up on the co-operation, co-ordination and
combined effects of several resources of knowledge, energy and time. Therefore I approach
this important matter of acknowledgement through these lines trying my best to give full
credits where it deserves.
I would like to take this opportunity to express my sincere thanks to my project supervisor,
Prof.Rajesh Singh who has been a great support from the beginning. Without his guidance
and encouragement this project could not have been completed. He was a constant source
of inspiration right from the beginning at each and every stage, from whom I learned loving
every piece of work and dedication towards work. He was ready to give a patient hearing
and give valuable suggestions in all the matters.
I am also thankful to Mr.Sukesh Kumar, Rohit Kumar and other members of lab for their
corporation and support. Present form of this project would not have materialized without
their kind support.
I cannot fail to thank all my dear friends and classmates who were always with me and
leading me towards my task.
I wish to convey my special thanks to staff of library and lab in-charges who were been so
co-operative during my project.
Lastly I thank my dear parents for their prayers, love and sacrifice to make a life for me.
NUTHAN PRASAD KB
5. 5
ABSTRACT
Intrusion detection in open field is of great value in many situations. Monitoring of remote
areas needs a lot of manpower and has its limitations. Seismic footstep detection based
systems for homeland security applications are an important additional layer to perimeter
protection and other security systems. This project is mainly concerned with the detection
of any human intrusion by the detection of the footsteps from a person from few tens of
meters away using an underground seismic sensor, Geophone.
Presence of footstep is indicated by the impulses in the geophone signal. Kurtosis, a
statistical measure is used to identify the impulses, can be applied for a short duration of
time for which a footstep exists. This method is less complex and computationally efficient.
Geophone signal is subjecting to a band pass filter, the output of band pass filter is read
through Analog to Digital converter of the microcontroller and stored in its memory at
every constant interval of time, which is determined by the frequency at which signal is
being sampled. Every constant number of values stored in memory, which are read through
microcontroller through ADC and stored in memory is subjected to kurtosis using
microcontroller. This forms a single sensor node. Many such nodes are connected in a
topology to build a Sensor Network.
Intrusion is indicated to control room when microcontroller of sensor node calculates
higher kurtosis value. Control room will send a unmanned vehicle to take action against the
intrusion.
Keywords: Intrusion, Footstep, Kurtosis, Sampling Frequency, Band Pass Filter, Analog to
Digital Converter, Sensor Node, Sensor Network.
6. 6
Contents
List of Figures......................................................................................................................................7
CHAPTER 1........................................................................................................................................9
1.1 Introduction ...............................................................................................................................9
1.2 Background................................................................................................................................9
1.3 Previous Work .........................................................................................................................10
1.4 Literature Review ....................................................................................................................14
1.5 Objective..................................................................................................................................14
CHAPTER 2......................................................................................................................................14
2.1 Designing the System ..............................................................................................................14
2.1.1 Sensor Network ................................................................................................................15
2.1.2 Control Room and Unmanned Vehicle.............................................................................16
2.1.3 Wireless Communication System.....................................................................................16
CHAPTER 3: Sensor Node and Intrusion Detection.........................................................................17
3.1 Geophone.................................................................................................................................17
3.2 Amplification...........................................................................................................................20
3.3 Band Pass Filtering..................................................................................................................21
3.4 Envelope Detection..................................................................................................................22
3.5 Kurtosis-Intrusion Detection ...................................................................................................26
3.5.1 Variation of Kurtosis with distance ..................................................................................27
3.5.2 Variation of Kurtosis with walking style..........................................................................28
3.5.3. Effect of Gain on Kurtosis and detection range...............................................................30
3.5.4 Variation of Kurtosis with number of walkers .................................................................31
3.6 Global Positioning System ......................................................................................................32
3.7 Transceiver ..............................................................................................................................34
3.8 Flowchart.................................................................................................................................35
CHAPTER 4: Control Room and Unmanned Vehicle ......................................................................36
4.1 Unmanned Ground Vehicle (UGV).........................................................................................36
4.2 Design and Construction of UGV............................................................................................37
4.2.1 Chassis..............................................................................................................................37
4.2.2 Motor ................................................................................................................................37
4.2.3 Power Supply....................................................................................................................38
4.2.4 Motor Driver Module .......................................................................................................38
4.2.5 Microcontroller.................................................................................................................40
7. 7
4.2.6 Global Positioning System................................................................................................40
4.2.7 Transceiver .......................................................................................................................41
4.3 Flowchart.................................................................................................................................42
CHAPTER 5: Wireless Communication System...............................................................................42
RESULTS..........................................................................................................................................44
Conclusion.........................................................................................................................................45
Future Work.......................................................................................................................................46
References .........................................................................................................................................47
APPENDIX I .....................................................................................................................................48
APPENDIX II- ARDUINO CODES.................................................................................................49
List of Figures
1.3.1 Experiment Setup………………………………………………………………………..….11
1.3.2 Footstep when sampled at 30 KHz……………………………………………………….....12
1.3.3 Geophone footstep data sampled at 1 KHz………………………………………………....12
1.3.4 Signal flow path………………………….…………………………………………………13
2.1.1 Design setup…………………………………………………………………………….…..15
2.1.2 Actual sensor node setup………………………………………………………………........15
3.1.1 Geophone sensor parts…………………………………………….………………………...17
3.1.2 Actual CD-24 geophone………………………………………………………………..…...18
3.1.3 Magnetic seismometer electric circuitry…………………………………………………….19
3.1.4 Frequency response of Geophone CD-24 ………………………….………………….…….19
3.2.1 Geophone Signal……………………………………………………………………………..20
3.2.2 Amplification Circuit…………………………………………………………………...........21
3.3.1 Band pass filtered geophone signal…………………………………………………………..21
3.3.2 Band pass filter…………………..…………………………….……………………………..22
3.3.3 Band pass filter equations…………..………………………………………………………..22
8. 8
3.4.1 Envelope detected signal of person running in 20m radius……………………………….…24
3.4.2 Envelope detected signal of person running in 5m radius…………………………………...24
3.4.2 Envelope Detector Circuit…………………………………………………………………...25
3.5.1 Footstep……………………………………………………………………………………....26
3.5.2 Variation of kurtosis as source walks towards the sensor…………………………………...28
3.5.3 Envelope signals of walking and running compared ………………..……....……..………..29
3.5.4 Kurtosis of walking and running compared …………………………………………………30
3.5.5 Kurtosis variation with gain………………………………………………………………….31
3.5.6 Variation of footstep data with number of walkers…………………………………………..32
3.5.7 Variation of average kurtosis with number of walkers………………………………………32
3.6.1 GPS module…………………………………………………….…………………………....33
3.7.1 433 MHz Transceiver………………………………………………………………………..34
3.8.1 Flowchart of sensor node…………………………………………………….……………....35
3.8.2 Flowchart of ISR for reading sensor data……………………………….…………………...36
4.2.1 Chassis……………………………………………………………………….………………37
4.2.2 150 RPM DC Motor………...………………..……………....……………………………...38
4.2.3 12 volt battery………………………………...……………………………………………..38
4.2.4 Motor driver module………………………………… ……………………………………..39
4.2.5 Arduino Mega 2650…………………………………………… …………………………...40
4.2.6 Parallax GPS module………………………………… …………………………………….41
4.2.7 433 MHz transceiver…………………………………………… ………………………….41
4.3.1 Flowchart of UGV………………………………… ………………………………………42
5.1.1 Wireless Communication Network…………………………………………………………43
5.1.2 433 MHz transceiver………………………………… …………………………………….44
6.1.1 No Intrusion Result Demonstration Setup………………………………………………….45
6.1.2 Intrusion Result Demonstration Setup ………………………………… ………………….45
9. 9
CHAPTER 1
1.1 Introduction
When a person or animal walks along the ground they emit seismic waves as a result of
the impact. These waves then propagate through the ground. A geophone is a sensor
that is able to measure the amplitude of these seismic waves in the ground. Geophones
can be used to measure everything from a car driving by to an outright earthquake. The
critical part is then establishing algorithms that allow the device to differentiate seismic
waves coming from a person versus other seismic activity.
This project will focus on a building a sensor node which is capable of sensing any
human intrusion by developing a simple yet effective algorithms to detect people.
Numerous algorithms exist for classifying data; however, most of them are intended for
continuous intrusion and are not computationally efficient. The algorithms utilized for
this project will need to be relatively simple so they can run in a low power
environment, but also still maintain their effectiveness in detecting people. Lastly, a
unmanned vehicle will be built to take action against any intrusion detected by sensor
nodes after the vehicle receives the command from the control room along with the
location information of sensor node.
The methodology behind the design of this system was to create a platform that later
could be built upon and expanded. The scope of the project is quite expansive and some
areas have had more attention than others. Every attempt has been made to clearly
document all the aspects of the project so that in the event that someone chooses to
pursue an aspect of the project at a later point the design and results will be at their
disposal.
1.2 Background
The ground, like any other elastic medium, allows waves to propagate through it [3].
The impact from a footstep hitting the ground can be distinguished from as far as 100
meters away under ideal conditions. The maximum distance is directly related to the
attenuation rate of the ground and the type of wave being studied. There are four types
of seismic waves that propagate through the ground: compression, shear, Rayleigh, and
Love. These waves have varying diminishing amplitudes as they travel through the
ground. The Rayleigh wave diminishes as 1/R while the shear and compression waves
10. 10
diminish as 1/R2. The Love wave, which is caused by the layering of the soil, is not
really considered. In footstep detection, the most important wave is the Rayleigh wave.
It is a wave which travels along the surface of the earth. Its components expand in two
dimensions and diminish exponentially with depth. As a result of this, the wave can be
detected at much further distances than the body waves (compressional and shear).
Another thing to consider is how the energy from a footstep gets partitioned into the
three waves. The shear and compressional waves which are body waves contain roughly
26% and 7% of the energy, respectively, while the Rayleigh wave contains 67% of the
energy. Therefore the Rayleigh wave is the critical wave for footstep detection. Not only
does it propagate through the ground over greater distances, but it is also where the bulk
of the energy from a footstep gets transmitted. It is also possible to extract bearing
information from the Rayleigh wave using a three-axis geophone.
1.3 Previous Work
There are numerous algorithms for classifying data; however, most of them require
massive amounts of computational processing power or memory. For instance, Succi et
al used a Levenberg-Marquardt Neural Network Classifier to track vehicle data [7]. This
produced good results, but it required 6MB of dynamic memory for its matrix
processing. In a low powered embedded system running a classifier of that nature would
require too much power.
Kenneth Houston and Dan McGaffigan at Draper Laboratory have done a significant
amount of research in the area of personnel detection using seismic sensors. Most
systems prior to their work were transient based. The downfall of that approach is that
many real-world signals unrelated to human locomotion look like transients. Systems
designed like that will have either a very high false alarm rate or else will be insensitive.
They introduced the idea of using spectrum analysis on envelope-detected seismic
signals. This method not only produced reasonable detection ranges but also was
significantly better at discriminating footsteps from other types of seismic sources.
But these algorithms assume that the footsteps are continuous and hence the spectrum
analysis can be applied, which is not really the case in terms of real life situations where
the intruder keeps a couple of steps and stop momentarily hence on spectrum analysis
the event is not detected. However Succi et al introduced a new technique of detecting
and differentiating the footstep from noise and other events. Footstep seismic signals are
different from other seismic signals by their impulsive nature. Measure of impulsive
11. 11
nature of any signal is called as kurtosis. Hence, kurtosis can be performed on seismic to
distinguish footstep seismic signal from other seismic signals.
Fig 1.3.1 shows the experimental setup of experiment conducted by Succi to record data
from one geophone sensor. The geophone is an external passive sensor with a spike on it
that penetrates into the ground. It generates very small electrical voltages depending on
the intensity of the propagating waves in the axis that the geophone is arranged in. When
the signal arrives at the DAQ board it is initially filtered through the analog circuitry. It
is filtered through two pole analog low pass filter. This filter band limits the signal to
prevent aliasing. The signal is captured using a 16 bit NI-DAQ at sampling frequency of
10 KHz through the NI LabView software installed on a Laptop.
Figure 1.3.1: Experiment Setup
13. 13
Figure 1.3.3 Geophone footstep data sampled at 1 KHz
The clearest way to understand how the system processes data is to analyze it from a
signal perspective. The next couple paragraphs describe the path a signal takes from
when the sensor picks up vibrations all the way to the determination of whether or not
the signal is classified as a person.
The signal processing chain is depicted in the Fig 1.3.4 . The Digital signal processing
begins by passing the data through a digital band pass filter allowing signals within the
range of 10 Hz to 100Hz to pass. This band pass range is geology dependent but was
fixed for this project. The next step is critical for the detection process. An absolute
value is performed on the whole signal to create an envelope of the received signal,
which will peak for each distinct footstep. The signal is processed in window of samples
for duration of 200ms. The kurtosis value of the window is calculated. This processing is
applied to data of the reference geophone only. The window size is chosen to satisfy all
possible separations between two footsteps so that none of the footsteps are missed.
Figure 1.3.4 Signal flow path
The value of the kurtosis is higher whenever impulsive events are present and below the
threshold of 3 when it’s absent. Footstep signals are characterized by instant raise in
magnitude of the geophone output signal when foot makes contact with the ground. This
character can be observed in the Fig 1.3.3. Thus if the value of kurtosis is greater than 3
it means an intrusion has occurred.
14. 14
1.4 Literature Review
Succi et.al work on kurtosis for the detection of footsteps was studied thoroughly and the
values of the kurtosis was recorded under various situations with variations in styles of
walking, recording environment(soil), noise environments, multiple walkers and
distance from the sensor[1]. The algorithm was written in such manner that the intrusion
is not missed in any of the cases if within range of the sensor. The algorithm was aimed
being as simple as possible but yet efficient enough to do the task.
Joe C. Chen et al.[2] have described at length in their paper how the seismic waves are
different from acoustic waves, difficulties in detecting and localizing signals in near
field compared to far-field and use of closed-form localization in order to localize the
footstep in near field.
1.5 Objective
Ultimately the goal of the project is to develop Intrusion Detecting System that are
computationally efficient in detecting any intrusion by human and communicate the
intrusion occurrence to the control room for taking action against the intrusion by
sending unmanned vehicle at the site of intrusion. In order to accomplish this, first a
sensor node has to be developed. Then, several such sensor nodes are installed in the
field where security is the concern. Secondly, a wireless communication system to
communicate between sensor node and control room and vice versa. Finally, unmanned
vehicle is developed to navigate to the intrusion sight and take defined necessary action
against the intrusion.
CHAPTER 2
2.1 Designing the System
Intrusion detecting system mainly has three sub systems. They are sensor network for
detecting human intrusion, control room and unmanned vehicle to take action against
intrusion, wireless communication system to communicate between sensor nodes,
control room and unmanned vehicle or robot.
15. 15
Figure 2.1.1: Design Setup
2.1.1 Sensor Network
Sensor network is a cluster of several sensor nodes arranged in a topology, best suited
for the system. Before designing the sensor network, sensor node has to be designed. In
this project more effort will be put on to design a sensor node by implementing the
theoretical concepts shown in Fig 1.3.1 which is altered and shown below to suit the
hardware implementation as in Fig 2.1.2
Figure 2.1.2 Actual Sensor Node Setup
Controller
and
Localizer
Node 1
Node 3
Node 4
Node 2
Alerting (Buzzer and Display
Device showing the location
of intruder)
Unmanned
Vehicle
16. 16
Fig 1.3.2 describes simulation of sensor node construction in MATLAB software. Here,
geophone data is recorded using data acquisition device and processed using MATLAB
simulation software. Whereas for hardware implementation, geophone signal is band
limited between 10 Hz and 100 Hz using a band pass filter. Envelope Detector tracks the
peak amplitude of the band limited geophone signal. This signal is subjected to kurtosis
operation by Microcontroller by reading Envelope Detector output through Analog To
Digital Converter. Intrusion is detected when kurtosis value is evaluated to be greater
than 3 by Microcontroller. When intrusion is detected, Microcontroller informs the
intrusion information to control room wirelessly. Information consists of time of
intrusion and location information like latitude and longitude of sensor node collected by
the GPS system interfaced with the microcontroller of sensor node. Otherwise, it keeps
looking for intrusion.
2.1.2 Control Room and Unmanned Vehicle
Any intrusion detected by sensor node is communicated to control room wirelessly.
Control room can initiate action against intrusion. Initiation could be taken by sending a
unmanned vehicle to the sensor node with special features like camera recording,
automatic weapon firing or it could be like turning on the camera installed at the
intrusion detected sensor node. Later, initiation will help in monitoring a specific
camera, which needs more attention than many other cameras in the vast field. However,
in this prototype project, only unmanned terrain vehicle is build, which navigates near
sensor node after detection of intrusion. Detection of intrusion is informed by sensor
node to unmanned terrain vehicle using 433 MHz Transceiver in this prototype project.
However, higher range wireless communication can be utilized based on the area of the
field under supervision.
2.1.3 Wireless Communication System
In this system, wireless communication can be implemented mainly between three
points. Firstly, it is implemented between sensor node and control room. Secondly, it is
used between control room and unmanned terrain vehicle. Lastly, wireless
communication is required between sensor nodes for future optimization of the sensor
network.
17. 17
In this prototype project, sensor node communicates using 433 MHz Transceiver, with
unmanned terrain vehicle bypassing control room. Also, wireless communication
devices with higher range are preferred over 433 MHz in real world.
CHAPTER 3: Sensor Node and Intrusion Detection
3.1 Geophone
A geophone is a small instrument for measuring ground motion. It is part of a category
of sensors called seismometers. Seismometers usually consist of a mass suspended by a
spring, with the mass being either a magnet that moves within a moving coil, or a coil
moving within the field of a fixed magnet. The geophone is an electromagnetic
seismometer, which produces a voltage across the coil that is proportional to the velocity
of the coil in the magnetic field, and thus approximately proportional to the velocity of
the ground. There are two variants of geophones a one-axis version and a three-axis
version. As the names imply, the one-axis version measures surface waves travelling in
one axis while the three-axis version provides circuitry to measure propagating waves in
all three axes. For this application only a single axis geophone is used. Fig 3.1 shows the
various parts of a geophone
Fig 3.1.1 Geophone sensor parts
18. 18
Figure 3.1.2 Actual CD-24 Geophone
For sensing motion, the simplest transducer is electromagnetic device, having coils moving
in the field of permanent magnet. The suspension or motion of mass suspended along
spring in electromagnetic field are caused due to seismic vibrations. This phenomenon
induces voltage in the coil. As a result current flowing in the coil exerts force opposing the
motion of the mass according to Len’s Law. This effect is used to damp the free suspension
of magnet motion.
The magnitude of the geophone output is given in volts and it is determined by the
conversion mechanism and a sensitivity factor. The voltage output of a geophone is directly
proportional to the product of velocity of the coil and its sensitivity factor. The geophone’s
sensitivity depends upon the number of loops in the coil and the strength of the magnetic
field of the permanent magnet. Figure 3.1.3 shows the electric circuitry of a geophone.
19. 19
Fig 3.1.3 Magnetic Seismometer Electric Circuitry
Resonance frequency of the geophone must be low in order to sense low frequency signals.
At the same time, it must also have larger bandwidth to sense larger frequency signals.
Every geophone is characterized by it frequency response. The natural frequency of the
geophone is in few Hz’s. The CD-24 geophone used in this project has a natural frequency
of 10 Hz. The Frequency response of the CD-24 sensor is shown below.
Fig 3.1.4 Frequency response of Geophone CD-24
0.1
1
10
100
1000
1. 10. 100. 1000.
Sensitivity(V/m/s)
Frequency (Hz)
20. 20
3.2 Amplification
Geophone outputs the voltage signal in few millivolts which is very small to supply into
band pass filter, envelope detector and microcontroller. Sample geophone signal is as
shown in the Figure 3.2.1.
Fig 3.2.1 Geophone Signal
From inspection, the average peak amplitude of the geophone signal is around 5
millivolts. But, Microcontrollers generally reads analog voltages from 0 volts to 5 volts.
Therefore, signal from geophone has to be amplified from 5 millivolts to 5 volts. Hence,
an amplifier of gain 1000 has to be used and the amplifier must be capable of amplifying
the small voltages in range of millivolts. Simple operational amplifier will not be
capable to amplify analog signal of very small strength. Also, simple operational
amplifier will not have the gain in the range of 1000. However, instrumentation
amplifiers like PGA202, PGA203 and AD623 can amplify such low amplitude signal
and also give gain up to 1000. Fig 3.2.2 shows the amplification connection using
AD623 instrumentation amplifier. From datasheet of AD623, the value of RG is given by
100 kilo Ohms / (G -1). Where, G is the gain value of AD623. We require a gain of 1000
to amplify 5 millivolts to 5 volts. Substituting the value of 1000 for G in RG = 100 kilo
Ohms / (G - 1), we get RG = 100.1001 Ohms. Signal from geophone is applied across the
pins 3 and 2. Output signal which is the amplified signal of the input is obtained across
the pin 6.
21. 21
Fig 3.2.2 Amplification Circuit
3.3 Band Pass Filtering
Generally, most of the energy of footstep signal is concentrated between the frequency
10 Hz to 100 Hz. A portion of the filtered data is show in Fig 3.3.1
Fig 3.3.1 Band pass filtered geophone signal
Signal from geophone is passed to band pass filter through internal amplifier. Band pass
filter is the combination of a low pass filter and a high pass filter. Signal is passed to high
filter with 10 Hz cutoff frequency to allow signal having frequency above 10 Hz, followed
by passing the output of high pass filter to a low pass filter having 100 Hz cutoff frequency
to band limit the signal between 10 Hz and 100 Hz. Design of high pass filter and low pass
filter with cutoff frequency 10 Hz and 100 Hz are as below.
22. 22
Fig 3.3.2 Band pass filter
Figure 3.3.3 Band Pass Filter Equations
Resonant or Central frequency, fr is taken as 50 Hz, low pass filter cut off frequency, fl is
taken as100 Hz and high pass filter cut off, fh is taken as 10 Hz.
Therefore, Bandwidth is calculated as fl – fh = 90 Hz, Hence QBP = 50/90 = 0.5555.
Substituting these values in the above equations, the following simplified equations are
obtained, R1 = R2*1.2345 and C1*C2 = 4.8970751*10^-11. Taking R1 = 1000 Ohm and C1
= 47 micro Farad, R2 and C2 are calculated and approximated as 1300 Ohm and 1 micro
Farad.
3.4 Envelope Detection
After the signal has been filtered it looks like the signal shown in Figure 3.3.1. Notice
how the impact from the footsteps generates both positive and negative waves. The
frequency of these waves is controlled by the terrain. Performing kurtosis on the entire
signal instead of just the peak amplitude of signal might generate wrong kurtosis value.
The goal however is to treat this whole block as one impact and then determine the
23. 23
parameters. This is achieved by tracking the peak amplitude of the signal and subjecting
it to the kurtosis operation. Tracking of peak amplitude is accomplished by the process
of envelope detection. Two methods were considered to implement the process. First
method, Hilbert Transform method is more suitable when the envelope detection process
is implemented by programming. Second method implements the same using a RC
hardware, diode and opamp.
In Hilbert Transform method, the concept of analytic signal or pre-envelope of a signal
x[n] can be described by the expression:
y[n]=x[n]+jx[n];
The envelope of the signal is given as
e[n]= + [n]);
The envelope is formed by taking the absolute value of the analytical signal. This
generates an envelope for each of the footsteps shown in Figure 3.4.
24. 24
Fig 3.4.1 Envelope detected signal of person running in 20m radius
Fig 3.4.2 Envelope detected signal of person running in 5m radius
25. 25
Hilbert Transform method is comparatively easier and efficient to implement, when it is
implemented by programming using microcontroller. But, microcontroller is busy in
reading the analog signal through Analog to Digital Converter and performs kurtosis
operation on the digital data. Second method uses opamp, diode, capacitor and resistor to
implement envelope detection. Hence, there is no extra burden on microcontroller of the
sensor node. Therefore, second method is used to perform envelope detection on the band
passed signal.
In second method, opamp is used to separate band pass filter and envelope detector. It also
compensates the voltage drop across diode. Diode allows only the positive part of the signal
to pass through it. Capacitor in the circuit gets charged, if the voltage due to storage of
charge is lesser then the voltage applied across it. Otherwise, it discharges through the
resistor. All these components put together as shown in the envelope detector circuit Fig
3.4.3 follows the envelope of the input signal at the output.
Studies have shown that the minimum amount of time taken for each step by human is 200
ms. Therefore, maximum number of steps taken in a second is 5. Hence, maximum value
for fm is 5 Hz. Geophone signal is band limited between 10 Hz to 100 Hz whose frequency
is represent as fc (10 Hz < fc < 100 Hz). These values of fc and fm, let the value of
time constant, ζ to be 0.125 which is the product of resistance RL and CL. Let CL
= 220 micro Farads. Then RL = 568.1818 Ohms, which can be approximated to
570 Ohms. Diode 1N4148 blocks negative part of the analog signal and Voltage
of 0.6 V drops across the diode. This voltage drop is compensated by Opamp,
LM325/NS.
Fig 3.4.3 Envelope Detector circuit
This signal is then supplied to microcontroller to perform the kurtosis operation on the
sampled signal, read through the analog pin of the microcontroller. Kurtosis is ratio of
26. 26
the 4th
to 2nd
moment of the amplitude of the signal. The kurtosis value is found to be
less for sinusoidal and Gaussian distributed signals and the kurtosis value increases with
the increase in impulsiveness of the signal.
3.5 Kurtosis-Intrusion Detection
Footstep signal is characterized with impulsive nature in its seismic signals periodically
with each impulsive part of the signal representing a footstep. Such impulsive nature can
be used to distinguish the seismic signals generated from footsteps from other seismic
signals generated by other events like vehicle movement, wind noise. Measure of
impulsiveness is called as kurtosis. Kurtosis calculated on footstep signal is found to be
greater than 3 due to its impulsiveness. Whereas, for non-impulsive seismic signal
possess kurtosis less than 3. Hence, performing kurtosis on seismic signal will help in
distinguishing footstep seismic signal from other seismic signals. Fig 3.5.1 shows the
seismic signal of a footstep.
Fig 3.5.1 Footstep
Kurtosis is ratio of the 4th
and 2nd
moment of the amplitude distribution of the signal.
Kurtosis varies only with the shape of the signal but not the amplitude. This concept is
very important for the technique to infer that kurtosis performed on walking and running
seismic signal do not differ.
For N samples of the seismic signal, the kurtosis is calculated:
Where µ is the mean over N samples;
27. 27
Kurtosis is implemented in sensor node by a microcontroller. For building a prototype
sensor node, Arduino Mega 2560 is utilized because arduino microcontrollers provide
greater flexibility and ease while coding and has many inbuilt functionalities like analog
to digital convertor. Microcontroller reads envelop detected signal through its analog
pin, stores in EEPROM of the microcontroller continuously for every few seconds,
depending on the sampling rate. Sampling rate is taken as 1000 Hz in this project, which
is sufficient enough to sample the envelope detected signal. Simultaneously, while
saving the data every 1/Sampling Frequency seconds, by using the concept of interrupt,
kurtosis operation is also performed by microcontroller.
The kurtosis can be calculated on the output of Band Pass Filter. But, this method of
calculating might lead to wrong determination of kurtosis as the entire signal is taken
into consideration. Therefore, performing kurtosis, after subjecting the band limited
signal through envelope detector would give more accurate kurtosis value as envelope
detector keep track of the peaks of the band limited signal. Kurtosis value computed on
sinusoidal signals and Gaussian noise signals are lesser than 2 and 3 respectively. Hence,
our technique is to analyze the envelope signal every 200 ms increments, and compute
the kurtosis of the signal in that sample. The reason for choice of 200ms as the window
time can be explained as follows. The maximum speed at which a person can run for a
short duration is 10 m/s. For covering 10m in a second the person has to take at least 5
footsteps 2 m apart in order to achieve it. Hence there can be 5 footsteps in a second at
max and the time interval between two successive impulses is a minimum of 200ms in
case of a single source. The cadence frequency of a human is defined as the number of
footsteps per second when he is walking normally. The cadence frequency turns out to
be 2Hz implying that 2 footsteps can be found 500 ms apart under normal condition.
The duration of the footstep is around 100-150ms. The impulse generated by the heel
striking the ground lasts for 50ms with smaller peaks due to the friction of the front part
of the ground for the remaining duration of the footstep.
3.5.1 Variation of Kurtosis with distance
The value of kurtosis varies with distance between source and sensor. When the source
is nearer to the sensor the value of kurtosis during the footstep is larger and when the
source moves away from the sensor the kurtosis value of the footstep decreases. At a
certain distance the kurtosis value falls below the threshold of 3. The range of the sensor
28. 28
is thus limited by the noise in the environment. When the surroundings are calmer then
the range of detection is larger than usual. It was observed that the detection range
considerably increased during night.
Figure 3.5.2 Variation of kurtosis as source walks towards the sensor
3.5.2 Variation of Kurtosis with walking style
Running footstep signal have higher impulsive amplitude compared to walking footstep
signal. But, impact of footstep on ground in running footstep signal is for shorter
duration comparatively. This difference can be noticed in the Fig 3.5.3
30. 30
3.5.4 Kurtosis values of running and walking compared
Though running footstep signal have higher impulsive amplitudes, they do not have any
advantages while detection compared to walking footstep signal because running
footstep signal are available in detection area for shorter duration. Therefore, detection
capability and ability on running seismic signal is not increased compared with walking
seismic signal. This argument is supported by the Fig 3.5.4
3.5.3. Effect of Gain on Kurtosis and detection range
It is desirable to increase the range of detection of the sensor in all circumstances. We
did investigate the effect of additional gain on the detection range of the sensor. This
was done by carrying out a simple experiment. Data was recorded without any gain
stage at first and the kurtosis value was calculated. The variation of kurtosis with
distance was noted. The detection range was determined by the distance at which the
kurtosis value falls below the threshold. Now the experiment was repeated by a gain of
10. The kurtosis value was calculated for this. It was observed that the gain had no effect
on the detection range of the sensor. The signal had same kurtosis values with and
31. 31
without the gain. This is illustrated by the figure 3.5.5 which shows the kurtosis value of
a footstep taken a same distance with and without gain.
Figure 3.5.5 Kurtosis variation with gain
3.5.4 Variation of Kurtosis with number of walkers
In case of single walker the value of kurtosis is directly proportional to the distance
between sensor and person. The closer person gets to the sensor larger is the kurtosis
value. Kurtosis gradually decreased with distance. The variation is already shown in
figure 3.5.2.
When the number of walkers was increased to 2 the value of the kurtosis fell to 5 from 8
at a distance of 7m. With increase in number of walkers the kurtosis value further
decreased. The reason for this was indistinctive impulses due to increased number of
walkers. Table 3.4.4 shows the variation of kurtosis at 7m with number of walkers.
32. 32
Figure 3.5.6 Variation of footstep data with number of walkers
Figure 3.5.7 Variation of Average Kurtosis with number of walker
3.6 Global Positioning System
The Global Positioning System (GPS) provides location and time information, anywhere
on or near the Earth using GPS satellites.
33. 33
When microcontroller calculates kurtosis on the data read though Analog to Digital
Convertor and stored in its EEPROM continuously. When it finds that the calculated
kurtosis value is greater than 3, it infers that the intrusion has taken place. At this point
microcontroller initiates GPS module to gather information like time of intrusion and
location information like latitude, longitude. This information is communicated to
control room or Unmanned Ground Vehicle wirelessly.
Fig 3.6.1 GPS Module
34. 34
3.7 Transceiver
Communication between sensor node, control room and UGV happens wirelessly
through transceivers connected with each block. Depending on the range, power
constraints, different wireless communication and transceivers can be utilized. However,
in this project 433 MHz Transceivers are used.
Fig 3.7.1 433 MHz Transceiver
36. 36
Fig 3.8.2 Flowchart of Interrupt Subroutine for reading sensor data
CHAPTER 4: Control Room and Unmanned Vehicle
4.1 Unmanned Ground Vehicle (UGV)
An Unmanned Ground vehicle (UGV) is a robot used to augment human capability in
both civic and military activities in open terrain. It is used as a human replacement in
several military operations such as capturing video, handling explosives, diffusing
bombs and front linear reconnaissance.
In this project, UGV receives command from control room to take action against the
intrusion, when sensor node detects intrusion. Sensor node informs its location attributes
37. 37
read from Global Positioning System (GPS) to control room; same is communicated to
UGV as well, which helps UGV in navigated to the sensor node, again with the help of
GPS.
4.2 Design and Construction of UGV
UGV in this project must be capable to move, to communicate with control room, and
navigate to sensor node using GPS and take action against the intrusion at sensor node.
To implement these functionalities, a chassis, four motors, power supply (12 V), motor
driver, microcontroller, GPS module are required.
4.2.1 Chassis
Chassis is a frame which structurally supports all the other parts of a vehicle, generally
made of metals. Fig 4.2.1 shows chassis supporting wheels and motors of UGV.
Figure 4.2.1 Chassis
4.2.2 Motor
In this project, four 150 rpm DC motors are used. Each motor drives a wheel of UGV. It
operated with 12 volt DC power supply. It is capable of rotating in both clockwise and
anticlockwise direction. However, polarities of the power supply have to be reversed, if
the motor has to be driven in anticlockwise direction. This property of motor is utilized
to drive the vehicle in reverse direction along with usual forward direction.
38. 38
Fig 4.2.2 150 rpm DC Motor
4.2.3 Power Supply
Dc motor chosen for the project operates at 12 volts to drive at 150 rpm. A 12 volt
power supply has to be shared by all the four motors, motor driver module and
microcontroller. Though DC motor operates at 12 volts, power is supplied to motors
through a motor driver module because motor driver module controls the polarity of the
voltage applied across the DC motor depending on direction it has to rotate.
Microcontroller operates at 5 volt supply. Therefore, 12 volts supply is regulated to 5
volts using LM7805 regulator chip to power microcontroller.
Figure 4.2.3 12 Volt Battery
4.2.4 Motor Driver Module
Motor Driver Module receives 12 Volts as input supply and applies across the DC Motor
with certain polarity depending on the requirement of the vehicle motion. Function of
39. 39
this module is to vary the polarities of the voltage applied across the motor depending on
the requirement of vehicle motion. The required direction of vehicle is conveyed to input
pins of the motor driver by microcontroller. It utilizes L298 integrated chip to implement
the functionality of controlling the polarity of the voltage applied across DC motor.
Motor Driver Module also has inbuilt LM7805, 5 Volt voltage regulator integrated chip,
which takes 12 Volts as the input and regulates it to 5 Volts which can be used to power
microcontroller.
Fig 4.2.4 Motor Driver Module
40. 40
4.2.5 Microcontroller
Role of microcontroller in this project is to wait for receiving the information of
intrusion from either control room or from a sensor node wirelessly. Information
received from sensor node, about the intrusion, consists of time of intrusion and location
information like latitude and longitude of the sensor node. This information is
communicated wirelessly to microcontroller of UGV. Microcontroller using Global
Positioning System receives its location information which helps it in knowing latitude,
longitude of its location. Knowing the location information of itself and sensor node,
microcontroller generates signals to motor driver module, to navigate near intrusion
detected sensor node and take action against intrusion.
Figure 4.2.5 Arduino Mega 2650
4.2.6 Global Positioning System
The Global Positioning System (GPS) provides location and time information, anywhere
on or near the Earth using GPS satellites.
GPS module installed in UGV receives its location information from GPS satellites.
Knowing location information of UGV, microcontroller in UGV, compares latitude,
longitude of UGV with latitude, longitude of sensor nodes and generates signals to
navigate towards that sensor node.
41. 41
Fig 4.2.6 Parallax GPS Module
4.2.7 Transceiver
Communication between sensor node, control room and UGV happens wirelessly
through transceivers connected with each block. Depending on the range, power
constraints, different wireless communication and transceivers can be utilized. However,
in this project 433 MHz Transceivers are used.
Figure 4.2.7 433 MHz Transceiver
42. 42
4.3 Flowchart
Fig 4.3.1 Flowchart of UGV
CHAPTER 5: Wireless Communication System
Wireless sensor networks (WSNs) provides area surveillance and informs intrusion, if
any to control room rapidly through wireless communication for rapid and flexible
43. 43
deployment of UGV. This scenario is shown in Fig 5.1.Black circular block represent
sensor node and hexagonal block represent control room. Sensor nodes after detecting
intrusion inform control room wirelessly. Followed by this, Control room communicates
with UGVs wirelessly. Then, UGVs navigate to sensor node where intrusion is detected
using GPS system.
Figure 5.1.1 Wireless Communication Network
Depending on the range, area of field of security, power constraints, environment different
wireless communication can be chosen. Wireless communication takes between sensor
nodes, control rooms and UGVs. Communication could take place between two sensor
nodes or between two control rooms could as well. Communication could be simplex or
duplex. All these factors depend on the design of wireless communication networks design.
However, in this project 433 MHz transceivers are used.
44. 44
Figure 5.1.2 433 MHz Transceiver
RESULTS
Geophone available with us is not sensitive enough to capture the seismic signals generated
by human footstep. Hence, the project is demonstrated using Signal Generator and DAC in
Arduino Due microcontroller board.
A sinusoidal signal of 5 millivolt amplitude is generated using Signal Generator. This
signal is corresponds to the non-footstep signal captured by geophone. If the frequency of
the signal is not in between 10 Hz and 100 Hz, then the signal is attenuated at band pass
filter of sensor node, hence no signal is available to process further in sensor node. This is
due to the reason that the sensor node is designed to work only with footstep signal, which
can have frequency between 10 Hz to 100 Hz. If the signal is confined between 10 Hz and
100 Hz, then the signal is amplified, band limited and subjected to envelope detector
circuits before subjecting to microcontroller for performing kurtosis operation. The signal
generated is sinusoidal in nature, which is different from the footstep signal generated in
geophone. Kurtosis calculated by microcontroller is always less than 3. This demonstrates
that the intrusion has not occurred. Hence Unmanned Ground Vehicle remains in its place.
This demonstration set up is as shown in the Fig 6.1.1
45. 45
Figure 6.1.1 No Intrusion Result Demonstration Setup
Generation of footstep signal, captured by geophone using signal generator alone is
difficult. However, Generation of the signal which is similar to the output signal of
Envelope Detector, when sensor node is subjected to footstep signal is easy. This signal is
generated using DAC of Arduino Due microcontroller board. This signal is applied to
microcontroller of sensor node for performing kurtosis on signal. Microcontroller of sensor
node finds that the kurtosis value is greater than 3, which indicates intrusion. This
information is informed to UGV wirelessly. Unmanned Ground Vehicle navigates towards
sensor node and stops.
Figure 6.1.2 Intrusion Result Demonstration Setup
Hence, all the scenarios are demonstrated using Signal Generator and DAC of Arduino
Due.
Conclusion
Result demonstrated by generating signal similar to geophone signal using signal generator
and DAC of Arduino Due microcontroller board shows that the geophone with sufficient
46. 46
sensitivity can be used to develop sensor node that can detect human footstep, thereby
detecting intrusion. However, the results of the project is demonstrated by generating the
signal similar to the geophone signals using Signal Generator and DAC of Arduino Due
microcontroller board which hold true with results that can be produced using geophone
with sufficient sensitivity too.
433 MHz transmitter and receiver are used to communicate between sensor node and UGV,
bypassing control room. As only one sensor node is developed, control room has been
bypassed. Using 433 MHz transmitter and receiver, maximum distance of communication
between sensor node and UGV is few meters. This distance can be increased by using
better form of wireless communication.
Topology of the sensor network can be altered according to the requirements of the area of
security. In this project one to one topology is used between sensor node and UGV
bypassing control room.
Navigation of UGV from control room to sensor node and vice versa has to be updated with
better technology. Obstacle avoidance of the UGV is also a concern.
Future Work
This project has huge scope for updating. Following are few updates that can be considered
as updates.
1. The Band pass filter used for filtering of footstep is fixed for this project. The
filter has to be made adaptive so that the filter can be modified based on the site
and incoming data.
2. Envelope Detection, Amplification processes has also need to be adaptive.
3. Unmanned All-Terrain vehicle, Quad copter can be used instead of simple UGV
with better technologies like obstacle avoidance, transmitting video wirelessly
and better algorithm to reach destination or intrusion detected sensor node with
shortest distance from control room.
4. Range of wireless transmission can be increased using better wireless
technologies like Wi-Fi, Zig bee, etc.
47. 47
References
1. Footstep Detection and Tracking, George Succi, Daniel Clapp, Robert Gampert,
Gervasio Prado.
2. Source Localization and Beamforming, Joe C. Chen, Kung Yao , and Ralph Hudson.
3. Problems in Seismic Detection and Tracking ,Dr. George Succi, Dr.Gervasio Prado,
Robert Gampert, Torstein Pedersen and Hardave Dhaliwal
4. Seismic Footstep Signal Characterization, A. Pakhomov, A. Sicignano, M. Sandy,
and T. Goldburt
5. Seismic Signals and Noise Assessment for Foot Step Detection Range Estimation in
Different Environments, Alex Pakhomov and Tim Goldburt
6. Personnel tracking using seismic sensors, Michael S. Richman, Douglas S. Deadrick,
Robert J. Nation, Scott L. Whitney
7. Demonstration System for a Low-Power Seismic Detector and Classifier, Elliot
Richard Ranger , Master’s thesis ,MIT 2003
8. The Generalized Correlation Method for Estimation of Time Delay , Charles H.
Knapp
49. 49
APPENDIX II- ARDUINO CODES
Code for Sensor node
#include <EEPROM.h>
#include <VirtualWire.h>
const int led_pin = 11;
const int transmit_pin = 12;
const int receive_pin = 2;
const int transmit_en_pin = 3;
int signalGenerator = A0; // select the input pin for the potentiometer
int const samplingFrequency = 1000;
int const lastEEPROMAddress = 1000;
int const sampleIncrement = 100;
int const delayBetweenSamplesMicro =
(int)(((float)(1/(float)samplingFrequency))*((((10*10)*10)*10)*10)*10);
int const windowSize = samplingFrequency/5;
int signalGeneratorValue = 0; // variable to store the value coming from the sensor
int addr = 0;
int i = 0;
float mean = 0.0;
float summationNeu = 0.0;
float summationDen = 0.0;
int minAdd = 0;
int maxAdd = minAdd + windowSize;
float kurtosis = 0.0;
float x = 0.0;
50. 50
float xx = 0.0;
float samples[windowSize];
void setup() {
TCCR0A = 0;// set entire TCCR0A register to 0
TCCR0B = 0;// same for TCCR0B
TCNT0 = 0;//initialize counter value to 0
// set compare match register for 2khz increments
OCR0A = ((16*10^6) / (samplingFrequency*64)) - 1; //(must be <256)
// turn on CTC mode
TCCR0A |= (1 << WGM01);
// Set CS01 and CS00 bits for 64 prescaler
TCCR0B |= (1 << CS01) | (1 << CS00);
// enable timer compare interrupt
//TIMSK0 |= (1 << OCIE0A);
TIMSK0 = 0x01;
sei(); // MAKE SURE GLOBAL INTERRUPTS ARE ENABLED
vw_set_tx_pin(transmit_pin);
vw_set_rx_pin(receive_pin);
vw_set_ptt_pin(transmit_en_pin);
vw_set_ptt_inverted(true); // Required for DR3100
vw_setup(2000); // Bits per sec
}
ISR(TIMER0_COMPA_vect){
signalGeneratorValue = (int)((float)analogRead(signalGenerator))/4.0;
EEPROM.write(addr, signalGeneratorValue);
addr++;
if (addr == 1000) addr = 0;
51. 51
}
void loop() {
while (maxAdd <= lastEEPROMAddress)
{
float samples[windowSize];
mean = 0;
for ( i = minAdd; i < maxAdd ; i++)
{
x = EEPROM.read(i);
samples[i-minAdd] = x;
mean += x;
}
mean = mean/windowSize;
summationNeu = 0.0;
for ( i = 0; i < windowSize ; i++)
{
xx = (samples[i] - mean);
xx = xx*xx;
xx = xx*xx;
summationNeu += xx;
}
summationNeu /= (float)(windowSize -1);
summationDen = 0.0;
for ( i = 0; i < windowSize ; i++)
{
52. 52
xx = (samples[i] - mean);
xx = xx*xx;
summationDen += xx;
}
summationDen /= (float)(windowSize -1);
summationDen = summationDen*summationDen;
kurtosis = 0.0;
kurtosis =(float) (summationNeu/summationDen);
if (kurtosis <3)
{
TIMSK0 = 0x00; // DISABLE OVERFLOW INTERRUPT (TOIE1)
Serial.begin(9600);
for ( i = 0; i < windowSize ; i++)
{
Serial.println(samples[i]);
}
Serial.println("");
Serial.print("Sample Frequency ");
Serial.println(samplingFrequency);
Serial.print("Min Address ");
Serial.println(minAdd);
Serial.print("Max Address ");
Serial.println(maxAdd);
Serial.print("Signal Generator Value ");
Serial.println((int)((float)analogRead(signalGenerator))/4.0);
Serial.print("Kurtosis ");
Serial.println(kurtosis);
53. 53
char msg[] = {'i','n','t','r','u'};
digitalWrite(led_pin, HIGH); // Flash a light to show transmitting
digitalWrite(transmit_en_pin, HIGH);
vw_send((uint8_t *)msg, 5);
vw_wait_tx(); // Wait until the whole message is gone
digitalWrite(led_pin, LOW);
delay(1000);
TIMSK0 = 0x01; // RE-ENABLE OVERFLOW INTERRUPT (TOIE0)
}
minAdd += sampleIncrement;
maxAdd += sampleIncrement;
}
minAdd = 0;
maxAdd = minAdd + windowSize;
}
Code for UGV
#include <VirtualWire.h>
const int led_pin = 6;
const int transmit_pin = 12;
const int receive_pin = 11;
const int transmit_en_pin = 3;
const int vcc = 8;
54. 54
int motorLeftForward = 22;
int motorLeftReverse = 23;
int motorRightForward = 24;
int motorRightReverse = 25;
int motorLeftEnable = 36;
int motorRightEnable = 37;
boolean permit = false;
boolean reach = false;
void setup() {
delay(1000);
Serial.begin(9600); // Debugging only
Serial.println("setup");
// Initialise the IO and ISR
vw_set_tx_pin(transmit_pin);
vw_set_rx_pin(receive_pin);
vw_set_ptt_pin(transmit_en_pin);
vw_set_ptt_inverted(true); // Required for DR3100
vw_setup(2000); // Bits per sec
vw_rx_start(); // Start the receiver PLL running
pinMode(vcc, OUTPUT);
pinMode(motorLeftForward, OUTPUT);
pinMode(motorLeftReverse, OUTPUT);
pinMode(motorRightForward, OUTPUT);
pinMode(motorRightForward, OUTPUT);
55. 55
pinMode(motorLeftEnable, OUTPUT);
pinMode(motorRightEnable, OUTPUT);
digitalWrite(vcc, HIGH);
}
void loop() {
digitalWrite(vcc, HIGH);
uint8_t buf[VW_MAX_MESSAGE_LEN];
uint8_t buflen = VW_MAX_MESSAGE_LEN;
//Serial.println("inside ");
if (vw_get_message(buf, &buflen)) // Non-blocking
{
char msg[] = {'i','n','t','r','u'};
int i;
digitalWrite(led_pin, HIGH); // Flash a light to show received good message
// Message with a good checksum received, print it.
Serial.print("Got: ");
permit = true;
for (i = 0; i < buflen; i++)
{
Serial.print(buf[i],HEX);
Serial.print(' ');
if ((char)buf[i] != msg[i])
{
permit = false;
Serial.print("Not Matching ");