Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Journal of Optical Communication Electronics

Volume 5 Issue 3

Raspberry PI Based Reader for Blind


Liya George1*, Rinsila Salim1, Riya Baby1, Anitta Thomas2
1
Student, 2Assistant Professor
Department of Electronics and Communication Engineering, Viswajyothi College of Engineering and
Technology, Vazhakulam, Kerala, India
Email: *liyageorgealucka@gmail.com
DOI: http://doi.org/10.5281/zenodo.3570357

Abstract
This project is an automatic document reader for visually impaired people, developed on the
Raspberry Pi processor board. It controls the peripherals like camera, a speaker which act as
an interface between the system and the user. Optical character recognition (OCR)
technology is used for the identification of the printed characters using image sensing devices
and computer programming. The OCR process is done using online and offline methods. It
converts images of typed or printed text into machine-encoded text. These encoded texts are
then converted into the audio output (Speech).Raspberry Pi is used for the translation of
printed document into data files using Tesseract library and Python programming. These
data files are then computed by OpenCV library and Python programming language to get
the audio output.

Keywords: Google cloud vision, optical character recognition, OpenCV, tesseract

INTRODUCTION people. OCR and Text-to-Speech synthesis


There are many existing solutions to the is used to convert images into audio output
problem of assisting individuals who are (Speech). The proposed apparatus has a
blind to read, however, none of them camera which act as the input device for
provide an efficient reading. We focus on digitization and this digitized script is
improving the competence of blind people processed by OCR (software module). A
by providing them with a solution where procedure is followed for recognition of
the details are given in the form of audio characters and the line of reading. In the
signal. Raspberry Pi-Based Reader is an context of software development, the Open
automatic document reader for visually CV (Open source Computer Vision)
impaired people using OCR technology. libraries are employed to capture image of
The proposed project uses a camera-based text and character recognition. The final
assistive device which can be used by identified text document is given to the
individuals to read printed text. The output devices based on the choice of the
scheme is to implement an embedded user. Headset connected to the Raspberry
system based image capturing technique Pi or a speaker act as the output device.
using Raspberry Pi board. The design is
inspired by prior research with visually LITERATURE SURVEY
impaired people, and it is small and Ray Kurzweil [1] proposes K-Reader
portable, that helps in achieving result in Mobiles, specially designed movable
little setup. Here, we have put forward a reading assistant for visually impaired
text read out system for visually impaired people. "K-Reader Mobile" runs on cell

11 Page 11-16 © MAT Journals 2019. All Rights Reserved


Journal of Optical Communication Electronics
Volume 5 Issue 3

phone and allows the user to read mail and blind people to aim the letters accurately,
many other documents. But this fails to finger-worn device assists the visually
give an economical solution. impaired in reading the paper-printed text.
Vasanthi G [8] proposes a Vision Based
Athira Panicker [2] proposes a smart Assistive System for Label Detection with
shopping assistant label reading system Voice Output. This camera-based assistive
with voice output for the blind using the
system help blind persons read text labels
raspberry pi. This system only reads
documents from clear flat surface and it and product packaging from a hand-held
does not read from complex backgrounds. object.

Marut Tripathi [3] proposes a Navigation BLOCK DIAGRAM OF PROPOSED


System for blind people to navigate safely METHOD
and quickly. In the system, ultrasonic The proposed project is implemented on
sensors and USB camera are used to detect the raspberry pi 3 board. The raspberry pi
and recognize obstacles. Once the 3 is a mini computer, it consists of 4 USB
obstacles detected via ultrasonic sensors, it ports, 40 GPIO pins for input or output,
sends feedback in the form of beep sound
CSI camera interface, full HDMI port, DSI
through an earphone to inform the person
about the hindrance. display interface, SOC (system on a chip),
LAN controller, micro SD card slot,
Dimitrios Dakopoulos [4] proposes a Bluetooth 4.1, audio jack, and video
Wearable Obstacle Avoidance Electronic socket, 5V micro USB connector and
Travel Aids for Blind that presents a Ethernet port.
comparative survey among portable or
wearable obstacle detection or avoidance Fig. 1 shows the block diagram of
systems to aware users about the progress raspberry pi based blind reader. It consists
in assistive technology for blind people. of raspberry pi 3, battery, camera, press
button, headset. Raspberry Pi is the control
X. Chen [5] proposes Automatic Detection
unit which controls the entire system. A
and Recognition of signs from natural
scenes, here they presented with an battery of 9V is use as the power source
approach for automatic detection and which turns on camera for capturing the
identification of signs from natural scenes image. Press button is used to activate the
and it's given to a sign translation task. The program and headset for audio output. The
system fails because it works only in the printed text is to be placed under the
Chinese language. camera to ensure the image of good quality
and fewer distortions. Then an applicable
William A Ainsworth [6] proposes a blind-assistive system, an algorithm might
system for converting English text into prefer. It checks the availability of all the
speech using an inexpensive computer but devices and also for the connection, when
it is not suitable for all memory range of the applications starts. Raspberry Pi 3 uses
computers as only small amount of stored a Linux based operating system named
data has scrutinized. Raspbian. The first part is booting the
Raspberry Pi board by installing the
Michael McEnancy [7] Finger Reader is an Operating system Raspbian OS and
audio reading gadget for Index Finger. installing the essential libraries and
Even Though this system doesn't help packages.

12 Page 11-16 © MAT Journals 2019. All Rights Reserved


Journal of Optical Communication Electronics
Volume 5 Issue 3

Figure 1: Block diagram of the proposed system.

Next is the image acquisition system, in on a chip), LAN controller, micro SD card
which we have interfaced a camera, to slot, Bluetooth 4.1, audio jack, and video
capture the image of the text document. socket, 5V micro USB connector and
The data is transfer to the OCR algorithm Ethernet port, power supply. Power Supply
which converts the image data to text data. Unit supplies electrical energy to the
The OCR algorithm scans the image and output loads. In real time the camera feeds
checks each alphabet or letter. Then it its images to a computer or computer
gives a corresponding text output after network, often via USB, Ethernet or Wi-
Fi. The Raspberry Pi board is connected to
verifying it with its database. We can use a
the Projectors, Monitors and TV through a
dictionary to compare the words detected
HDMI to VGA Converter.
by the algorithm for auto-correction. We
have chosen text to speech engine which Raspbian being a free operating system
convert the text data to an audio output and based on Debian developed for the
is plays through the earphones connected Raspberry Pi module. The operating
to the audio jack. system is a set of basic programs and
services that helps the Raspberry Pi run.
ARCHITECTURE OF THE Many versions of Raspbian are available
PROPOSED SYSTEM like Raspbian Stretch and Raspbian Jessie.
The proposed system has the hardware and As of the latest update Raspbian uses
software part. PIXEL, Pi Improved X-Window
Environment, and Lightweight as its
Hardware Implementation fundamental desktop environment.
The hardware parts of the system include
the Raspberry Pi Battery, Camera, Press Python 3.7
Button, and Headset.

Raspberry Pi
It is a credit card sized minicomputer that
plugs into a computer monitor or TV and it
uses standard keyboard and mouse.
Raspberry Pi 2 and Raspberry Pi 3 are the
2 models of Raspberry Pi. The hardware
components of the Raspberry Pi include 4
USB ports,40 GPIO pins for input or
output, CSI camera interface, full HDMI
port, DSI display interface, SOC (system Figure 2: Raspberry Pi.

13 Page 11-16 © MAT Journals 2019. All Rights Reserved


Journal of Optical Communication Electronics
Volume 5 Issue 3

Battery Software Implementation


Python is a high-level programming This includes the programming part of the
language for general-purpose. Design project.
ideology of python emphasizes on code
readability, notably using significant Raspbian
whitespace. It provides clear programming Open CV
on both large and small scales. Python OpenCV is an open source library for
features a dynamic type system and programs and functions mainly aimed at
automatic management. It supports real-time computer vision. Initially
multiple programming paradigms, developed by Intel, and later taken up by
including object-oriented, procedural, and Willow Garage then Itseez and again
has a large and comprehensive standard acquired back by Intel. Under open-source
library. license it’s free for the users and the
library is cross platform. OpenCV supports
9V 1.2Ah Battery is use for power supply. the deep learning frameworks TensorFlow,
The Larson Electronics BATT-9V-10PK is Torch/PyTorch and Caffe.
a ten pack of 9V 1.2Ah batteries. These
high-energy batteries feature up to 10 Flite
times more than carbon zinc battery and up Speech synthesis is the imitation of human
to 5 times more energy than alkaline voice. A computer system used for this
battery. This is a long lasting battery, task is called a speech synthesizer. This
because of the high running voltage, low synthesizer is accessible to all in hardware
impedance, better performance at low or software products. The foremost aim of
temperatures, wide operating temperature Text-To-Speech system is to translate
range and advanced safety features. This is normal language text into audio.
a lightweight battery which is alternative Synthesized speech can be produced by
to any comparable 9V battery. concatenating pieces of recorded speech
that are stored in a database. A system
Press Button which stores phones or diaphones provides
A press-button or simply button is a simple the largest output range, but this may give
switch mechanism for controlling some low clarity. For specific application
aspect of a machine or a process. Press domains, the storage of entire words or
button is used to activate the program and sentences allows for high-quality output.
headset for audio output. Press Buttons are Alternatively, a synthesizer can constitute
typically made out of very hard materials a model of the vocal tract and other human
like plastic or metal. The surface of button voice characteristics to create a fully
is usually flat or shaped to accommodate synthetic voice output. The quality of a
the human finger. So it can be easily
speech synthesizer is decided by its
depressed or pushed.
naturalness or similarity to the human
Headset voice and by its ability to be understood
A headset combines a headphone with a clearly.
microphone. Headsets are setup with either
a single-earpiece (mono) or a double- FLOW OF PROCESS
earpiece (mono to both ears or stereo).The Image Capturing
audio output is plays through the Image capturing is the first step. It is
earphones connected to the audio jack. achieved by using the press button
Headsets is also provide the equivalent interfaced to the Pi. To improve the quality
functionality of a telephone handset but of the image high resolution camera is
with hand free operation. used.

14 Page 11-16 © MAT Journals 2019. All Rights Reserved


Journal of Optical Communication Electronics
Volume 5 Issue 3

Dictionary
The recognized characters are cross-
checked with the database provided. The
database used for the online process is
taken from the cloud library whereas for
offline we use a trained dataset.

Image to Text Conversion


Here, the image is converted to machine-
encoded text. In the online process, we use
the Google cloud vision as the platform for
the conversion where Application Program
Interface (API) is used.

Figure 3: Flow of process. Speech Synthesis


This module performs the task of
done using Flite algorithm. The final audio conversion of the transformed machine-
output is heard using a headset. encoded text to the audible form. It is here,
we represented a system to scan written
RESULT text, for helping the blind individuals.
Image is captured and processed using Word recognition on the text regions is
OCR technology and Google Cloud performed using OCR. For this
Vision. Text to speech conversion is done methodology the camera acts as the input.
using flite algorithm. As the Raspberry PI board is high powered
it makes the camera streaming. The image
Character Recognition is captured by pressing the button when
The captured image is first enhanced, and the item for text reading is positioned
character recognition is done either by ahead of the camera. The image is
online or offline methods. In the offline converted to the document using Tesseract
method, Tesseract library and Python library. Text-to-Speech synthesis is used to
programming are used. Here the text files pronounce the document through the ear
are processed by various libraries like phones.
OpenCV, NumPy. In the online method,
Google Cloud Vision is used. Additional features can be added to the
system for obstacle detection using
CONCLUSION ultrasonic sensors.

REFERENCES
1. Ray Kurzweil (2008), “K-Reader
mobile user guide”, Knfb Reading
Technology Inc.
2. Ms. Athira Panicker, Ms. Anupama
Pandey, Ms.Vrunal Patil YTIET (Oct
2016), “Smart shopping assistant
label reading system with voice
output using Raspberry Pi”,
University of Mumbai, IJARCET,
Volume 5, Issue 10, ISSN: 2278-
Figure 4: Result.
1323.

15 Page 11-16 © MAT Journals 2019. All Rights Reserved


Journal of Optical Communication Electronics
Volume 5 Issue 3

3. Marut Tripathi, Manish Kumar, 1,


Vivek Kumar, Warsha Kandlikar (Jul- 6. William A Ainsworth (Jun 1973), “A
Aug 2014), “A navigation system for System for converting English text
blind people”, IJRASET, Volume 2, into Speech”, IEEE Transactions on
Issue 4, Audio and Electroacoustic, Volume
4. Dimitrios Dakopoulos, Nikolas G 21, Issue 3,
Bourbakis (Jan 2010), “Wearable 7. Michael McEnancy (July 2014),
obstacle avoidance electronic travel “Finger Reader is audio reading
aids for blind”, IEEE Transactions on gadget for Index Finger”, IJECCE,
systems, man and cybemetics, Volume 5, Issue 4,
Volume 40, Issue 1, 8. Vsanthi G, Ramesh Babu (Jan 2014),
5. Xilin Chen, Jie Yang, Jing Zhang, “Vision based assistive system for
Alex Waibel (Jan 2004), “Automatic label detection with voice output”,
detection and recognition of signs Dept. Of ECE, DMI College of
from natural sceneries transactions on Engineering, Chennai.
image processing”, Volume 13, Issue

16 Page 11-16 © MAT Journals 2019. All Rights Reserved

You might also like