CURRENCY DETECTOR AND VOICE CALLER FOR THE

BLIND USING ARTIFICIAL INTELLIGENCE AND

MACHINE LEARNING.
By

ODHIAMBO THOMAS ATIENO

Registration Number: SC212/1418/2017

A Research Proposal Submitted in Partial Fulfillment of the Requirements for the Award of the
Degree of Bachelor of Science in Software Engineering, Department of Computer Science,
School of Computing and Information Technology, Murang’a University of Technology.
 DECLARATION AND APPROVAL
I declare that this research is my original work and has not been submitted for any academic
award in any other university.

Signature……………………………. Date…………………………

Odhiambo Atieno.

SC212/1418/2017

This research proposal has been submitted for examination with my approval as the university
supervisor.

Signature…………………………… Date…………………………….

Peter Mwangi Maina

Lecturer Department of Computer Science

School of Computing and Information Technology

Murang’a University of Technology.

DEDICATION
This research proposal is dedicated to my mother Madam Jecinta and in memory of my father
Mr. Jeremiah Atieno for their love, support, and inspiration all through my academic journey. I
cannot forget about my siblings Kennedy Oluoch and my sisters Ruth Akinyi and Milka
Achien’g for their tremendous support which helped me sail though this academic journey.
ACKNOWLEDGEMENT
All form of praise and thanks is due to Almighty God, the creator of mankind, the most merciful and
gracious for his blessings, protections, courage and guidance.

I would like to acknowledge and thank my lecturer Mr. Peter Maina, who stood by me and made sure I
did everything regarding this project efficiently and effectively, my gratitude to him is unmatched. I also
thank my parents for their immense support, guidance and encouragement to make sure that I
complete this program successfully. I would also like to acknowledge my lecturers and all who have
contributed to this point in the success of my academic pursuit. I also thank my friend Gordon, Joseph
and Fred for their great support.

ABSTRACT
One of the major challenges facing visual impaired people is money recognition especially for paper
currency. In this paper we present a simplified system for currency recognition on Kenyan banknote. Our
proposed system is focused on image processing utilities that insure appearing the process as speedy
and as robust as possible. The basic techniques utilized in our proposed machine encompass image
foreground segmentation, histogram enhancement, area of interest (ROI) extraction and template
matching primarily based at the cross-correlation among the captured picture and our records set. The
experimental effects exhibit that the proposed method can understand Kenyan paper cash with
excessive first-rate reaches 89% and short time.

Once more, in this work an evaluation of speech technologies and their applications that provide or
augment access to the published or

Digital statistics, the day by day or social activities, as well as the personal or public centers for blind or
low vision persons is

Supplied. Speech technologies are presently taken into consideration to be vital for presenting trendy
reason interfacing besides offering accessibility for the folks that are visually impaired.
 CHAPTER ONE
1.1 Introduction
The capability to identify currency (including both coins and notes) without human manual
input is lacking in a great number of computer systems. Possibly the most critical one is helping
visually impaired people. In keeping with government

Kenya the variety of visually disabled people was determined to be higher. About 9% have been
estimated to be blind and around 18 percentage had low vision.

Current development of mobile systems makes the concept of currency recognition with a
smart phone an attractive one. In

This study we model an easy approach of template matching with SURF key point detector for
Android platform.

We are representing an app in which currency is recognized by means of app and end result is
sent through audio gadgets. One in every of the primary issues experienced by people with
visual impairment is the inability to perceive the paper currencies due to the approximation of
paper texture and length between different currencies. As a result, the role of this system is to
device a technique to solve this hassle which will make to make blind people experience safety
and determination within the financial approach.

There are two types in money recognition studies discipline; Scanner-based and camera-based.
Scanner-based systems are supposed to experiment the entire paper. Such systems are suitable
for the system of currency counters. At the same time Camera based totally structures besides
capturing the currency by means of a camera which can also seize a part of the currency. Most
related works in documentation assign with the scanner-primarily based type. For visual
disabled utilization, it’s assumed to enable the users to capture any part of the currency by the
use of their smartphones and let the device identify it and notify them of the currency value.

On this paper, digital camera-primarily based Kenyan currency is trained to be identified using
image processing, artificial intelligence and machine learning. The processing time of the data is
also very short and the system has the ability recognize currency captured in limited and high
contrast light conditions.

The Algorithm is meant for use by blind and visually impaired people who have problems in
currency recognition. We first attain a dataset of currency images taken by blind and normally
sighted people. From this dataset, we manually label and extract the text regions. Next, we
carry out statistical analysis of the text areas to decide which photograph features are
dependable indicators of textual content and have low entropy (i.e., Feature reaction is similar
for all text images).

We obtain vulnerable classifiers by using the usage of joint chances for function responses on
and of textual content. Those susceptible classifiers are used as enter to enhance machine
leaning algorithm to train a strong classifier.

1.2 Problem Statement

Over the years, the blind and the visually impaired have undergone a greater challenge when it
comes to dealing bank notes as they are not able to tell exactly what amount it is. This has led
to several instances of them being tricked and conned by those people with ill-intentions.
The following systems exist but the have not fully solved the problems that the blind undergo
daily:

1.2.1 The manual system used by the blind to tell the differences between different Kenyan
Bank notes: This method is not reliable especially when the bank notes get a bit older.
To determine the value of the currency, a person is required to touch the left and right
edges of the notes for raised bars. 1 bar denotes Ksh50, 2 bars Ksh100, 3bars Ksh200,
4bars Ksh500 and 5bars Ksh1000. Some of the blind people may also have sensitivity
issues with their thumbs because the bars are very small.
1.2.2 App Developed by Central Bank of Kenya to help Determine if a currency is fake.
The Central Bank of Kenya (CBK) recently launched a new smartphone application that
helps users familiarize themselves with the new currency features.
The app provides a way to detect fake currency in 4 different ways.
 Using Ultraviolet (UV) light.
 Using Normal Light.
 Using Feel/ Sense of touch.
 By tilting the note.

This app is useful in detecting fake currency; however, it cannot help the visually
impaired persons because there is no audio feedback communicated back to them.

1.2.3 Manual way by checking on the currency manually using light.

Portrait Watermark A three-dimensional portrait of a lion's head can be seen when the
note is held up to the light. This cannot be of any use to the visually impaired and the blind
because they will not be able to see.

1.2.4 Checking the serial numbers.

The serial numbering style is asymmetrical and has progressively larger digits in adjacent
positions. However, this cannot be seen by the visually disabled.
1.2.5 Conclusion.

Having closely examined the systems and models above for currency detection in Kenya, It has
become clear tome that there is a need for me to develop a more advanced system that will
solve the shortcomings of the above systems and models with a priority given the visually
disabled.

1.3 Objectives

The main objective of the project is to help the visually impaired and the blind to identify the
Kenyan currency notes automatically

and be informed of the currency value though audio means. This project will also help the users
to make calls by simply issuing a command through voice. Other model in existence though, this
model is designed to overcome the drawbacks of the previous methods. This model gives a
faster and more accurate output when compared to the others.

1.3.1 Specific Objectives

a) To gather and analyze the requirements that are needed to develop a currency
detector.
b) To gather and analyze the requirements that are required to develop a voice caller app.
c) To identify original currency note using image processing techniques.
d) To develop a system that can compare images of currency note to the stored images of
original currency note images.
e) To offer inexpensive and accurate system to the user which could without difficulty be
accessible and gives accurate recognition of currency notes.
f) To develop user friendly Mobile application for currency recognition system that gives
feedback by means of audio of the value of the currency.
g) To develop a user-friendly mobile application that allows users to make calls by means
of audio.
 h) To make available the system to visually disabled people quickly and easily so they can
utilize it anywhere and at any time.

Significance of the study

The results of the study will be of great significance to in the following ways:

The blind user: This study will help the blind user to be able to determine with ease the values
of the currency they are interacting with even if the currency is old and they no longer able the
feel the bars on the edges.

From this study, we are able to come up with an application that will help the blind users to
make phone calls just by the use of their voices.

With the ability of the system to differentiate between genuine and fake currencies, the blind
user can now be confident of the money he/she will be receiving.

Scope of Study

This project will consist of creating a friendly android application that will be used by visually
disabled people in Kenya.