Image Transmission over GSM Voice Channel

Final Year Project Report

Presented
by
Ramshaw Zafar
CIIT/SP11-BET-074/ISB

Sara Nasir
CIIT/SP11-BET-083/ISB

Saima Huma Rana

CIIT/SP11-BET-079/ISB

In Partial Fulfillment
of the Requirement for the Degree of

Bachelor of Science in Electrical (Telecommunication) Engineering

DEPARTMENT OF ELECTRICAL ENGINEERING

COMSATS INSTITUTE OF INFORMATION TECHNOLOGY,

ISLAMABAD
Jan 2015
 Image Transmission over GSM Voice Channel

Final Year Project Report

Presented
by
Ramshaw Zafar
CIIT/SP11-BET-074/ISB

Sara Nasir
CIIT/SP11-BET-083/ISB

Saima Huma Rana

CIIT/SP11-BET-079/ISB

In Partial Fulfillment
of the Requirement for the Degree of

Bachelor of Science in Electrical (Telecommunication) Engineering

DEPARTMENT OF ELECTRICAL ENGINEERING

COMSATS INSTITUTE OF INFORMATION

TECHNOLOGY, ISLAMABAD
Jan 2015
 Declaration
We, hereby declare that this project neither as a whole
nor as a part there of has been copied out from any
source. It is further declared that we have developed this
project and the accompanied report entirely on the basis
of our personal efforts made under the sincere guidance
of our supervisor. No portion of the work presented in
this report has been submitted in the support of any other
degree or qualification of this or any other University or
Institute of learning, if found we shall stand responsible.

Signature: ______________
Name: Ramshaw Zafar

Signature: ______________
Name: Sara Nasir

Signature: ______________
Name: Saima Huma Rana

i
 COMSATS INSTITUTE OF INFORMATION
TECHNOLOGY, ISLAMABAD
Jan 2015
Image Transmission over GSM Voice
Channel
An Undergraduate Final Year Project Report submitted to the
Department of
ELECTRICAL ENGINEERING

As a Partial Fulfillment for the award of Degree

Bachelor of Science in Electrical (Telecommunication) Engineering
by
Name Registration Number
Ramshaw Zafar CIIT/SP11-BET-074/ISB

Sara Nasir CIIT/SP11-BET-083/ISB

Saima Huma Rana CIIT/SP11-BET-079/ISB

Supervised by

Dr. M. Junaid Mughal

Professor,
Department Of Electrical Engineering
CIIT Islamabad

COMSATS INSTITUTE OF INFORMATION

TECHNOLOGY, ISLAMABAD
Jan 2015

ii
 Final Approval
This Project Titled
Image Transmission over GSM Voice Channel

Submitted for the Degree of

Bachelor of Science in Electrical (Telecommunication) Engineering
by
Name Registration Number
Ramshaw Zafar CIIT/SP11-BET-074/ISB

Sara Nasir CIIT/SP11-BET-083/ISB

Saima Huma Rana CIIT/SP11-BET-079/ISB

has been approved for

COMSATS INSTITUTE OF INFORMATION TECHNOLOGY,

ISLAMABAD

_____________________
Supervisor
Dr. M. Junaid Mughal,
Professor

______________________ ______________________
Internal Examiner-1 Internal Examiner-2
Dr. Adeel Israr, Mumtaz Ali,
Assistant Professor Lecturer

______________________
External Examiner
Dr. Shahzad Saleem,
Assistant Professor

_____________________
Dr. Fasih Uddin Butt
Department of Electrical Engineering

iii
 Dedication
Dedicated to our beloved parents, supporting teachers and our educational institution,
COMSATS Islamabad.

iv
 Acknowledgements

All praise for ALLAH ALMIGHTY who gave us courage and perseverance to take
this project to its completion.
We express our deepest gratitude from the core of our hearts to our parents whose
loving, caring, supporting and benevolent personalities have left an incredible
impression on our characters.
We especially acknowledge guidance of our supervisor Prof. Dr.M. Junaid Mughal
for his kind support at every turn during this two semester journey. He encouraged us
to believe in ourselves. He envisioned the unique idea of this project. Had he not
bestowed us with his professional experience and advices, this project would not have
been completed so efficiently.
We are also indebted to the vital supporting teachers of our institution, Lecturer
Nouman Naseer, Lecturer Raza ul Islam and Assistant professor Ahsan Khwaja who
guided us through several technical problems encountered during the project.

Ramshaw Zafar
Sara Nasir
Saima Huma Rana

v
 Table of Contents
1. Introduction 1

1.1 Fundamental Idea. . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Aims and Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.3 Problem Definition. . . . . . . . . . . . . . . . . . . . . . . . . . . .2

1.4 Block diagram. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . 2

1.5 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3

2. Literature review 4

2.1 AT Commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4
2.1.1 Software used. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4
2.1.2 Call dialing mechanism. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . 6
2.1.3 Interfacing AT Commands with GSM voice channel. . . . . . . . 7

2.2 Serial communication in MATLAB . . . . . . . . . . . . . . . .7

2.2.1 Test data transmission. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.2 Serial image transmission. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .9
2.2.3 Testing data with different Images. . . . . . . . . . . . . . . . . . . . . .10
2.2.4 Working with Grid in MATLAB. . . . . . . . . . . . . . . . . . . . . . .10
2.2.5 Working with Black and White Image . . . . . . . . . . . . . . . . . . 11
2.2.6 Working with Colored Images . . . . . . . . . . . . . . . . . . . . . . . . 11

2.3 Image Compression . . . . . . . . . . . . . . . . . . . . . . . . . .11

2.3.1 Discrete Cosine Transform . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3.2 Decreasing spatial resolution . . . . . . . . . . . . . . . . . . . . . . . . . .13
2.3.3 Reducing bits per pixel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3. Transmission over GSM Voice Channel 15

3.1 Hardware in the loop . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.2 Fundamental approach . . . . . . . . . . . . . . . . . . . . . . . 16

3.2.1 Uni-polar NRZ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . .17
3.2.2 Binary Frequency Shift Keying . . . . . . . . . . . . . . . . . . . . . . . .17
3.2.3 Fast Fourier Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2.4 Multi Frequency Shift Keying . . . . . . . . . . . . . . . . . . . . . . . . .19

vi
 3.4 GSM Channel Impairments . . . . . . . . . . . . . . . . . . . 20
3.4.1 Issues in the channel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.4.2 Filter Designing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.5 Channel Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . .21

3.6 System Testing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .22

3.7 Results of Real Time Implementation . . . . . . . . . . . . 23

4. Learning Outcomes 24

4.1 Future Recommendations . . . . . . . . . . . . . . . . . . . . .24

4.2 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

Appendix A 26
Bibliography 27

vii
 List of Acronyms
GSM……………………………………… Global System for Mobile Communication
MMS………………………………………….……… Multimedia Messaging Service
GPRS……………………………………………………General Packet Radio Service
MFSK……………………………………………… Multiple Frequency Shift Keying
DCT………………………………………….…………… Discrete Cosine Transform
IDCT……………………………………….………Inverse Discrete Cosine Transform
MATLAB………………………………………...…………………Matrix Laboratory
PC……………………………………………………….………… Personal Computer
DMM……………………………………………………………… Digital Multimeter
BFSK…………………………………………….…… Binary Frequency Shift Keying
FSK……………………………………………………………Frequency Shift Keying
DFT……………………………………………..…………Discrete Fourier Transform
IDFT………………………………………………Inverse Discrete Fourier Transform
FFT…………………………………………………………… Fast Fourier Transform
IFFT……………………………………………………Inverse Fast Fourier Transform
RGB……………………………………………..…………………… Red, green, blue
NRZ………………………………………………………………… Not return to Zero

viii
 List of Figures
1.1 Block diagram……………………………………………………………....3
2.1 HyperTerminal………………………………………………………….…...5
2.2 Steps for connection establishment…………………………………….…...5
2.3 Call dialing mechanism with AT commands…………………………….…6
2.4 Testing AT commands at HyperTerminal…………………………………..6
2.5 RS-232 female pin out………………………………………………………8
2.6 String of characters transmission………………………………………...…9
2.7 Serial transmission of black and white image……………………………...9
2.8 Serial transmission of colored image…………………………………….…9
2.9 Image compression steps…………………………………………………...11
2.10 DCT of true color image……………………………………………………13
2.11 Spatial resolution……………………………………………………………13
2.12 Reducing bits per pixel………………………………………………..……14
3.1 Hardware…………………………………………………………………….15
3.2 Pictorial description of hardware…………………………………….……..16
3.3 Real time transmission………………………………………………………17
3.4 (a) NRZ Uni polar digital signal…………………………………………….18
(b) BFSK modulation………………………………………………….…….18
3.5 (a) Fourier Transform of NRZ Uni polar signal………………………….…18
(b) Fourier of BFSK signal………………………………………………….18
3.6 (a) Digital signal……………………………………………………….…….20
(b) Modulated analog signal………………………………………………....20
3.7 Flow chart for channel testing……………………………………………….22
3.8 Image transmission results on GSM channel………………………………..23

ix
 Abstract
This project aims at designing a communication system that can transmit low
resolution images on GSM voice channels through the use of numerous techniques
such as M-FSK, image compression, DCT, quantization, encoding and decoding, in
the form of voice tones at ordinary call rates irrespective of availability of GPRS/3G
and broadband facilities. Although video and images are being transmitted nowadays
via GPRS and MMS, but it is worthwhile to transmit them in the form of data in audio
format over GSM voice channel. Doing so, data can be transferred to the receiver at
normal call rates. Hence, for those areas in which GPRS facilities are still not
available, it will be substantially beneficial to transmit video over the voice channel.
This project will be of significant importance to media industry.

x
 CHAPTER 1
INTRODUCTION
GSM (Global System for Mobile) is worldwide standard for cellular communication, with
availability in more than 219 countries and 90% share in the global market. GSM is an open,
digital cellular technology for transmitting mobile voice and data services. There is a broad range
of cellular phones that operate on GSM. Hence, consumers have greater flexibility in selecting a
handset that suits them, and are not limited to purchase cellular phones locally available in their
country.

Transmission of image over GSM channel is not a straight forward task. Firstly, the GSM
channel is band limited. Secondly, GSM would cause great delay as compared to voice channel
since GSM channel has not been designed for real time applications. These limitations put
forward a stumbling block in accomplishing this challenging idea. Although video and images
are being transmitted nowadays via GPRS (General Packet Radio Service) and MMS
(Multimedia Messaging Service), but it is worthwhile to transmit them in the form of data in
audio format over GSM voice channel. Doing so, data (in the form of audio signals) can be
transferred to the receiver at normal call rates. Hence, for those areas in which GPRS facilities
are still not available, it will be substantially beneficial to transmit video over the voice channel.

In this thesis, we will have a comprehensive discussion on the modulation schemes used for
successful transmission of image over GSM channel and how this technique will perform in
hardware and software. Finally the results of modulation schemes will be presented and
analyzed.

1.1 Fundamental Idea

The idea basically implies presenting cheap alternative for media vans (future extension of this
project). The integrated system would comprise a few mobile phones in parallel on the
transmitter and the receiver end. Compressed video (in the form of audio) will be transmitted and
the data will be recovered at the receiver end. Though the video will be of low quality and there
will be delay encountered at the receiver end but these shortcomings are trivial when compared

1
to the significant benefits this integrated system will provide to the news agencies and media.
There are certain remote areas where media vans cannot reach. GPRS facilities are also rarely
available in some remote areas of Pakistan. Hence this system will efficaciously overcome these
two limitations.

1.2 Aims and Objectives

This project aims at transmitting image in the form of voice tones at ordinary call rates
irrespective of availability of GPRS/3G (Third generation) and broadband facilities.

1.3 Problem Definition

One of the drawbacks of GSM channel is its bandwidth when working with projects such as
video transmission. GSM is band limited channel with its bandwidth ranging from 300 Hz to
3.5k Hz. The bandwidth of video is 6 M Hz. One solution to this problem is to compress the
subsequent images in video so that its bandwidth is between the desired ranges.

1.4 Block Diagram

To transfer a large data over band limited channel, it needs to be compressed. Therefore, for
transferring a three dimensional colored image, its each dimension is split and compressed. A
colored image has three channels, red, green and blue. Using compression technique image is
compressed in each dimension. An image is composed of pixels, each pixel specified by some
decimal value. These values are converted into binary. Polar line encoding scheme is applied on
binary data. Then after frequency shift keying modulation, image is converted into sound. This
audio signal is transferred on voice channel. Decoding algorithm is designed on the receiver side
for reconstruction of image. The block diagram gives pictorial view of above stated process.

2
 Figure 1.1 Block Diagram

1.5 Applications

Transmitting data over GSM channel is indeed beneficial for areas which lack GPRS facilities. If
this concept is put functional by an integrated system having few mobile phone on either side, it
can be much economical substitute for expensive media vans.

In developing country like Pakistan, providing expensive media vans to every reporter is not
possible. Due to high cost, media vans are available only in main cities of Pakistan. Replacing
them with a cheap alternative would lead to efficient coverage of news in main cities especially
in the areas that lack GPRS, 3G and other broadband facilities. Thus this project holds great
importance in media industry.

3
 CHAPTER 2
LITERATURE REVIEW
Since this project aims at transferring image over GSM voice channel, hence instead of dialing
the call from mobile again and again, we can introduce automatic call dialing and reception
mechanism by interfacing mobile with PC using AT Commands in MATLAB. Apart from GSM
modem, we can also control cell phone. Considering this, we have used AT commands.

2.1 AT Commands

AT is an acronym for ’attention’. AT commands are ‘attention commands’ which are used to
control GSM modems. Basically these are the commands which are fed to the modem to perform
wired or wireless dial-up functions.

Syntax: All commands start with ‘AT’. Different commands have been designated for different
purposes. The command for sending SMS message to the carrier is +CMGS. Similarly the
command for call dialing is +ATD.

2.1.1 Software used

Hyper terminal is a communication based program. It comes with windows operating system,
98 versions. This software enables dialing up connections with other PCs via internet using
internal modem. It can also be used as a tool for setting up modems. Using AT commands in the
hyper terminal windows, we can control cell phone with various commands i.e. sending text
messages and dialing calls.

Nokia PC Suite is software used for interfacing Nokia cell phones and computers that have MS
windows operating system.

HyperTerminal is built in software in most of the operating systems. It holds the key features to
use AT commands. For a windows operating system it can be found as shown in figure 2.1.

4
 Figure 2.1 HyperTerminal

We used Nokia 2700 classic which functions on frequency bands GSM 850 / 900 / 1800 / 1900.
Following steps were followed:

1. Install Nokia PC suite into computer system and connected phone to computer through a
data cable.
2. Select PC suite option through phone. On PC suite, Nokia phone will appear on the
monitor.
3. Click on phone. Go to its properties and see on which port it is connected.
4. Now open HyperTerminal window and follow the following steps.

Figure 2.2 Steps for initializing modem

5. Create a new File. Click on properties to change settings. ASCII Setup. Mark options of
Echo, append, force and wrap.

5
2.1.2 Call Dialing Mechanism
Initially AT is sent to the GSM modem for testing whether the connection is established or not. If
the modem returns OK response it means the connection is established otherwise error is
returned. A simple command for call dialing was tested. ATD followed by the destination
number and semicolon was written on HyperTerminal. The modem returned OK response on the
screen meaning that call has been made as shown in figure 2.3.

Figure 2.3 Call dialing mechanism with AT commands

Figure 2.4 Testing AT commands at HyperTerminal

6
2.1.3 Interfacing AT Commands with GSM Voice Channel

For modem control, GSM channel can be interfaced with MATLAB software using the serial
USB port of personal computers and laptops. This way the required control using AT commands
can be obtained.
Call dialing: ATD command performs the action of dial tone in voice mode. This command
establishes the voice call and data in the form of voice can be then sent.
Automatic call control: For data transmission, uninterrupted ongoing call to the receiver is
required so that the data transmission is not hanged up due to call drop. This will ensure
complete transfer of data. Using the ‘automatic dialing’ and ‘automatic answer’ command, the
uninterrupted video transmission is made possible. If the call is dropped meanwhile transfer of
data is taking place, we just need to press a single button for resumption of data transmission.
Call re-initiation: In case of call drop, the transmission of video is interrupted hence data is lost.
The ‘dial tone’ command is used. If there is any remaining data in the voice receiver buffer, a
dial tone is detected hence this tone is sent in voice receiver mode. This code is sent every three
seconds if the data terminal equipment does not react to this code.

2.2 Serial Communication in MATLAB

Prior to shifting on actual channel of transmission, we practiced data transmission on wired

medium. Using the procedure of serial communication, any image taken was converted into row
vector and data was sent serially, bit by bit. On reception side, data was received bit by bit.
Those bits were recombined to construct an image.
In telecommunication environment, serial communication is the process in which data is sent bit
by bit i.e., one bit at a time. Serial port is designed so that serial cable RS-232 can be plugged
into it. RS-232 cable provides slow speed serial communication.

Figure 2.5 RS-232 female pin out

7
Bi-directional Communication: The serial port in a PC can transmit and receive data at a same
time. It uses different lines for transmission and reception. However, some of the serial devices
can only communicate uni-directionally by using two wires only (one for transmission and other
for ground). Steps for serial communication of data are as follows:

1. First of all you have to create a serial port object for a particular serial port by using a "serial"
command. While creating the object we have to configure properties of serial port according to
our requirements. We have set the baud rate 9600 bps. ‘Terminator’ and ‘timeout’ parameters are
also defined in our serial code. Terminator indicates the end of message. Baud rate is defined as
number of bits per second. It is used for measuring speed in communication. For example, 400
baud rate means 400 bits per second.

2. Make a connection between the object and the device. We can open the port for the
transmission of data by using the command "fopen" , write any data by using the commands
"fprintf" or "fwrite" and read the data using "fscanf", "fread" or "fget1".

3. Now for closing the port i.e. disconnecting it from device we use the command "fclose" and
to delete it from the memory we use "delete".

2.2.1 Test Data Transmission

Using specific codes for serial transmission of data, first of all a string of characters was sent as a
test code. This string of asterisk was successfully received at the receiver end. Digits from 0 to 9
were also sent and received serially on the other PC.

Figure 2.6 Test data transmission screenshot

8
2.2.2 Serial Image Transmission

Following the same procedure, image was converted into bits. These bits were transmitted
serially and were received bit by bit respectively on the receiver end. The successive bits were
displayed in command window at the receiver end and subsequent image was reconstructed.

(a) (b)

Figure 2.7 (a) Image sent (b) Image received

Likewise, the results of serially transmitted colored image are follows:

(a) (b)

Figure 2.8 (a) Image sent (b) Image received

2.2.3 Testing Data with Different Images

Gray scale image: Color constituents of such images are only shades of gray. Color
pixels of such images require very less information. They take much less space in the
memory and can be useful for hard wares and systems which support only 8-bit images.

9
 Comparing with RGB image, a gray scale image can be described as one having red,
green and blue components of equal intensities.
Binary image: A digital image in which each pixel can have either 0 or 1 value is a
binary image. They can be easily compressed and take up much less space in the
memory.
Indexed image: General syntax for indexed image is:Image(X); colormap(map)
Map is a m-by-3 array containing values in the range [0, 1] therefore red, green and blue
components are mapped into a single color by each row.
RGB image: It consists of three channels namely red, green and blue. For 24-bit RGB
image, each channel is composed of 8-bit.
True color image: When a RGB image is displayed using 24-bit color depth, it makes up
a true color image.

2.2.4 Working with Grid in MATLAB

On the transmitter side, a 3 x 8 matrix grid was plotted in MATLAB. After performing NRZ
modulation, the matrix was converted into digital signal. The time period for each wave was set
so that the frequency of the resulting signal does not exceed 3.5 k Hz. This was important
because while working with the GSM channel, only the frequency ranging from 300 Hz to 3500
Hz is useable. Then FFT of digital signal was performed and samples were plot. Binary
modulation was performed on the subsequent digital signal. The samples for BFSK modulation
were dominant in the range 1000 Hz to 1500 Hz. The modulated signal was converted into
sound.

At the receiver side, envelop detection demodulation technique was performed. Subsequent
digital signal was converted into matrix form and matrix was then converted into image.

At this stage, the channel delay was not considered hence this modulation/demodulation scheme
involves simple techniques.

2.2.5 Working with Black and White Image

Black and white image was first converted into one dimensional image and same procedure was
opted for modulation/demodulation scheme.

10
2.2.6 Working with Colored Images

Dimensions of a 3D colored image are separated into three channels. Each channel is converted
into row matrix. Modulation/demodulation techniques are separately performed on each row
matrix to convert it into sound. At the end three channels of are combined at the receiver side
and image is reconstructed.

2.3 Image Compression

The goal of image compression is to reduce the amount of data required to represent a digital
image while preserving as much information as possible. It does not change the resolution of
image. Various steps of image compression are:

Figure 2.9 Image compression steps

2.3.1 Discrete Cosine Transform

Sequence of data points can be expressed as summation of cosine functions oscillating at

different frequencies. This is the underlying principle of DCT compression technique. Cosine
functions are crucial to compression since less number of cosine functions are used to
approximate signal. This fact makes DCT more adaptable in compression applications. DCT is
analogous to DFT but it only uses real numbers. In DCT technique, each pixel in the image is
replaced by sinusoidal wave whose frequency is dependent on the degree of visual factor offered
by each pixel. In the start of DCT matrix, more important pixels are placed which are represented

11
by lower frequency. Higher frequencies represent less important pixels and are placed at the end
of DCT matrix. Hence the pixels are redistributed in the order of their visual importance. The
pixels having low visual elements are discarded causing trivial effect on the image. The
reconstructed image is clearly discernible although it contains reduced number of DCT
coefficients. Hence, an image with less number of DCT coefficients can be transmitted in much
less time.

For reconstructing the image, inverse discrete cosine transform (IDCT) is taken. Then image is
up scaled to its original dimensions and pixel values are restored. Compromise has to be made in
loss of data to achieve high speed transmission. Amongst eight variants of DCT, the four are
common. The most commonly used variant is ‘DCT-II’ which is referred as the DCT and ‘DCT-
III’ which is the inverse of DCT-II. In actual practice, DCT-II is transpose of DCT-III and vice
versa. In practical applications, DCT-II is used in signal processing and image compression due
to its energy compaction property. DCT-IV variant is used in MP3 audio compression.

Fig. 2.10: DCT of true color image

12
2.3.2 Decreasing Spatial Resolution

Spatial refers to the number of independent pixel values per inch. Image consists of pixels. Pixels
are the smallest part of picture element. Resolution defines how much information is contained
within an image. Decreasing the resolution of image means decreasing the information contained
within an image.

Figure 2.11: Spatial resolution

2.3.3 Reducing Bits per Pixel

A true color 24 bit RGB image has three channels. Each channel has 8 bits per pixels i.e. each
pixel is represented by 8 bits. Image can be compressed by any desired factor to reduce the
number of bits per pixels. Compressing an image by factor of 2 convert 24 bit image to 12 bit
image. Hence each channel will now have 4 bits per pixels i.e. each pixel that was formerly
represented by 8 bits in binary is now represented by 4 bits. Then image is rescaled by the same
compression factor.

Figure 2.12 Reducing bits per pixel

13
 CHAPTER 3
TRANSMISSION OVER GSM VOICE CHANNEL
Our main development tool is MATLAB. This was chosen because it has several advantages
over other software’s when dealing with image and video processing. It has the ability to process
both still images and video efficiently and also we are very familiar with MATLAB. We found it
easy to interface our cell phones with GSM channel when working with AT commands.

3.1 Hardware in the Loop

In this case, the channel is not simulated by software but we are using real cell phones instead.
However the data transmitted from the cell phones is then processed in MATLAB. Only one
mobile was used as the sender and one mobile was used as a receiver. Mobiles were connected
with PC using simple jack. At the sender terminal, mobile was connected with a speaker jack of
the PC. At the receiver terminal, mobile was connected with a mic jack of the PC. Simple voice
call was made between two mobiles and the sound signal (converting image into sound) from the
sender terminal. That sound signal was received on the other side and was decoded accordingly.

Figure 3.1 Hardware

Microphone of headset was removed. Microphone is input device and readily picks up
surrounding noise. Disabling the microphone prevents surrounding noise from interfering with
the transmitted audio signal.

14
 Figure 3.2 Pictorial Description of Hardware

3.2 Fundamental Approach

Real time data transmission involves synchronizing the data for transmission (for example
during video conferencing or telephone call). This is contrary to time – shifting approach where
data is transmitted after being stored [2]. For real time transmission, flv format video requires
bandwidth of 1-3 M Hz whereas, the GSM channel, we are dealing with, has around 2k Hz
bandwidth. Hence, we either need to increase the channel bandwidth or split the video data into
smaller compressed clips so that it can be transmitted over limited bandwidth. Increasing the
channel bandwidth is not possible.

15
 Figure 3.3 Overview of fundamental approach

3.2.1 Uni polar NRZ

It is one of the forms of line coding which directly encodes the bit stream. In unipolar encoding
scheme, high voltage represents 1 bit. For 0 bit, no voltage is transmitted. Since there is no rest
condition, it either represents 1 or 0 hence it is called ‘not return to zero’.

3.2.2 Binary Frequency Shift Keying

Since the image pixel values have been converted to digital data, hence using BFSK, the digital
data was converted into analog signal. For example, separate frequencies are assigned to values 0
and 1 in FSK modulation. When this modulation is performed on digital data, the respective
frequency values corresponding to 1 and 0 appear in the final modulation.

16
3.2.3 Fast Fourier Transform

FFT is an efficient algorithm for computing DFT and IDFT. FFT breaks the required calculations
into smaller ones so they can be computed rapidly. In DFT, the number of calculations required
is (N * N) where N represents the number of data points. FFT approach reduces this (N * N) to
(N log N). The log is to base 2. Since rate of increase of log N is much less than N, hence
considerable time can be saved while working with FFT. This difference in speed can be
substantial especially while dealing with longer data where N may range in thousands.

FFT algorithms can be efficiently implemented in a wide range of mathematical operations

ranging from simple complex-numbers operations to group theory and number theory [6].

17
3.2.4 Multi Frequency Shift Keying

MFSK is much more helpful than FSK or any other modulation technique used for sound
transmission. 16FSK has been used here for the transmission of the data to the receiving end.
This technique uses more than two frequencies to represent any bit. MFSK systems generally
space the tones as closely as possible, to restrict the transmitted bandwidth. Following are the
steps followed using MFSK:

Modulation:

1. Image was taken in RGB scale.

2. Then double matrix was made to perform any kind of algebraic expression.
3. Then it was converted into binary vector.
4. Then a row matrix was made.
5. Then 48 was subtracted from row matrix because it was changed due to any kind of
operation we have applied above.
6. Then bunch of frequency for bits were made.
7. M was taken as M=16 so total frequencies will be 16, one frequency for each bit.
8. Then ‘compared matrix’ named ‘comp’ was generated for each bit.
Comp = [0 1 2 4 . . . 15].
9. Then trainee signal was created for avoiding the signal to b loosed if signal was delayed
on the receiving end.
10. A time vector was generated of 50 bits.
11. Then a loop was made to modulate the signal. Each bit was multiplied to its related
frequency.
12. After that signals were plotted.
13. Then trainee signal was added in the start and end of the Modulated signal.
14. At the end it was made a sound file and was transmitted to the receiving end.

Demodulation:

1. Initially bunch of carrier frequencies were made for each bit of the signal i.e. 16.
2. Then bit stream was made by multiplying carrier frequencies with received signal.

18
 3. Then integration of bit stream was taken for every 50 values of the signal because length
of time vector was 50.
4. Then signal was retrieved by picking up the maximum value after integration because
after the multiplication of carrier signal with the received signal the relative signal was
raised to the higher value as compared to others.
5. Then signal was converted to single row and was made binary vector.
6. After that it was made into 8 bit chunks because values of Image were in 8 bit chunks.
7. Then three frames were made for RGB Image (Red Green Blue).
8. At the end Image was displayed on the screen with title “Recovered Image”.

Modulated Signal
1.5

Magnitude 0.5

-0.5

-1

-1.5
0 50 100 150 200 250 300 350 400 450 500
Time Vector

Figure 3.6 MFSK graphs

3.4 GSM Channel Impairments

GSM is a technology that provides voice and data transmission services. GSM employs TDMA
(time division multiple access). When talking specifically about the voice channel, the bandwidth
is limited to 3000 Hz. Frequency range for voice transmission is 300 Hz to 3300 Hz whereas
frequency range for data transmission is 600 Hz to 3000 Hz.

3.4.1 Issues in the Channel

Two major issues that are faced during data transmission over GSM voice channel are:

1. Inherited delay: GSM voice is a multipath channel. The delay is inherited since there is
always some delay during call establishment. To establish connections, GSM requires 28-31
seconds. When a call is dialed, it transfers from one base station to next and so there is delay.

19
This delay is not easily perceivable during call dialing process but such a delay becomes
significantly prominent when data is transmitted over voice channel. To compensate for this
delay, a known sequence of samples known as ‘training signal’ is introduced at the transmitter
and the receiver end. This training signal is introduces at the start of data. So when a call is
established, this random data in training signal is lost during that delay period instead of actual
data. Hence original data is preserved.

2. Inherited noise: GSM voice channel is noisy. This noise is introduces at the start of data,
since no matter what we do, the ambient noise cannot be completely prevented. Secondly the
modulators and demodulators of cell phones and laptops or PCs also induce certain amount of
noise in the data at both ends which significantly degrades the quality of received data. Hence we
need to design appropriate filters at the receiver end so as to minimize the noise as much as
possible.

3.4.2 Filter Designing

In digital signal processing, there are two types of filters:

1. FIR (finite impulse response)

2. IIR (infinite impulse response)

A band pass FIR filter (500 Hz to 3500 Hz) is designed. Its impulse response is finite because
there is no feedback. FIR can be easily designed and are simpler to implement.

3.5 Channel Testing

Prior to sending actual video, the channel was first tested for noise and delay in the GSM
channel. This step was crucial since we need to know actual values of parameters including bit
period, amplitude, starting frequency and ending frequency which give nearly accurate results for
the received data. Since any image was first converted into bits for digitization then further
modulation/demodulation operations were performed on it so we started testing the channel with
single sinusoid and proceeded towards multiple sinusoid file on transmitter side generates a
single frequency which is played on speaker output of the PC. Mat file on receiver side receives

20
data for 15 seconds from PC mic when it is executed and then result is saved in mat file on
receiver side.

Following steps are followed:

Figure 3.7 Flow chart for Channel testing

Setup the whole system as shown in the figure at the bottom. Make a call from transmitter to
receiver. When the call is established, play the receiver code. Immediately play the transmitter
code. When the codes finish execution, received data will be saved in mat file. Analyze the data
to look for pattern. If the data is not correct, run the codes again by changing the start and stop
frequencies.

3.6 System Testing

Using cable as medium of transmission: Two laptops were used. MATLAB sound file was
played on transmitter laptop and recorded on the receiver laptop. The sound was played back.
There was no noise in the recorded sound. Hence the wire was working properly.

21
Using channel as medium of transmission: On transmitter side sound file was played. Call was
dialed from transmitter mobile to receiver mobile. When connection was established, sound was
heard on other side but there was background noise along with transmitted sound.

3.7 Results of Real Time Implementation

Initially a small test image was taken for this purpose. Dimensions of image are 16 x 16
(width=16 pixels, height=16 pixels). Bit depth is 32 pixels. Bit depth quantifies how many
unique colors are available in an image's color palette in terms of the number of 0's and 1's, or
"bits," which are used to specify each color. However, one may take any image. Large sized
images can be compressed.

Using MFSK

Parameters of reconstructed image:

Mean square error (MSE) = 2.2720 x 104, PSNR (Peak signal to noise ratio) = 4.56

Figure 3.8 Results of real time transmission

22
 CHAPTER 4
LEARNING OUTCOMES
Channel is unpredictable. A simulation at same frequency and amplitude gives different results
every time. Inherited delay in channel: Training signal which is known to both transmitter and
receiver end should be introduced at the start of data. It will prevent actual data loss during the
time needed for call establishment. It thus ensures correct reception of data at the receiver side.

Channel noise is unavoidable: Appropriate filter designing (band pass filter) should be designed
to reduce noise as much as possible.

4.1 Future Recommendations

We followed the procedure by converting image pixels into digital and then performed MFSK on
the image. This is, however, an inefficient technique for transmitting image. Rather than
converting image pixels into digital, each pixel value should be assigned different frequency.
Hence for 256 pixels there will be 256 different frequencies. MFSK reduces the bandwidth
which makes it an efficient technique for transferring images and videos over band limited
channels.

Project extension: After successful image transmission, this project can be used video
transmission. For this purpose, few mobile phones in parallel can be connected on transmission
and reception side to transmit high quality images in less time. One image will be sent from each
mobile. The frames will be reconstructed on receiver side to construct video again.

4.2 Summary

In this age of Mobile Communication, images/videos are being sent anywhere using
GPRS/MMS and other internet facilities. To transfer data without using GPRS over GSM
Channel is a unique idea. Through our project, it is proved that data can be transferred on a voice
call at the normal call rate.

Basically the whole idea is based on replacing the media coverage vans with a small system that
would have some mobile phones which will be transferring data in parallel to receiving mobiles.

23
In this way the data will be retrieved at the receiver and one would be getting the replica of what
he has actually sent.

Our approach for solving this problem was that the video/image was captured and encoded in
Matlab, then created sound of this data in order to make it travel on voice channels, then the data
would be split according to number of mobiles available to us and then by dialing simple calls
from each sender to its particular receiver, those encoded sounds were transferred.

At the receiver, the collected data was rearranged in its original order and was being decoded to
get the actual information.

24
 Appendix A

References
http://www.ehow.com/list 6322410 advantages-disadvantages-gsm.html

http://www.businessdictionary.com/definition/real-time.html

http://searchnetworking.techtarget.com/definition/MFSK

http://www.academia.edu/4230927/An_Introduction_To_Digital_Image_Processing_ With_Matlab

http://photographylife.com/why-downsampling-an-image-reduces-noise

http://en.wikipedia.org/wiki/Fast_Fourier_Transform

http://www.ee.surrey.ac.uk/Research/PortfolioPartnership/I-labProjects/project1.html

http://www.zoltrix.com/support_html/PUBLIC/MODEM/ATmanual/ATHSP56.HTM

25
 Bibliography
[1] G. Adiv, “Determining Three-Dimensional Motion and Structure from Optical Flow Generated by
Several Moving Objects,” Vol. PAMI-7, pp. 384-401, July 1985.

[2] ETSI. Digital cellular telecommunications system (phase 2+); enhanced full rate (efr) speech
transcoding; (gsm 06.60 version 8.0. 1 release 1999). Technical report, ETSI, 1999.

[3] ETSI. Digital cellular telecommunications system (phase 2+); voice activity detector (vad) for
enhance full rate (efr) speech tra_c channels; (gsm 06.82 version 8.0. 1 release 1999). Technical
report, ETSI 2000.

[4] ETSI. Digital cellular telecommunications system (phase 2+); channel coding (3gpp ts 05.03
version 8.9.0 release 1999). Technical report, ETSI, 2005.

[5] ETSI. Etsi publications download area. http://pda.etsi.org/pda/queryform.asp,2008

[6] IPdrum. Ipdrum mobile skype cable. http://www.ipdrum.com,2008

[7] N.Katugampala, S.Villette, and A.M. Kondoz. Secure voice over gsm and other low bit rate
sytsems. Technical report, Center for Communication Systems Research, 2003.

26