B.M.

S COLLEGE OF ENGINEERING
Bull Temple Road, Basavanagudi, Bangalore - 560 019,

OPTICAL TEXT TO
SPEECH CONVERTER
USING OCR AND TTS

Guided By :-
DR. LALAITHA.S
 INTRODUCTION

• The Optical text-to-speech (TTS) converter is a transformative assistive technology

designed to empower individuals by enabling them to access textual information in
images and translate to different languages.

• Languages are the oldest way of communication between human beings whether
they are in spoken or written forms.

• Therefore, we have started to digitize these images, extract and interpret the data by
using specific techniques, and then perform text-to-speech synthesis (TTS).

• It is done in order to read the information aloud for the benefit and ease of the user.
Text extraction and TTS can be utilized together to help people with reading
disabilities

• This project has represented the innovative idea as well as a low cost technique that
is used to hear the contents of the text image without reading them.
 LITERATURE SURVEY

S TITLE IEE LIMITATION

L

N
O
 PROBLEM DEFINITION

• The problem is to translate the language for the individuals who have language
barrier.

• Even for travelers or tourists who travel to different states for vacation or business
can have language barriers.

• The individuals have reading disabilities.

• The lack of accessible, efficient, and natural-sounding TTS solutions limits the
usability of systems for visually impaired individuals, language learners, and smart
device users.
 PROPOSED SOLUTION

• An Optical Image Text-to-Speech (OITS) translator is a system that

automatically converts printed or handwritten text into speech.

• This system integrates Optical Character Recognition (OCR) to extract text

from images (such as photos of documents or books) and Text-to-Speech (TTS)
synthesis to vocalize the extracted text.

• Text-to-Speech Conversion: The extracted text is converted into speech using

the Festival Text-to-Speech (TTS) engine. The system can read the text aloud in
realtime with a clear and natural voice.

• User Interaction: The system is designed for ease of use, requiring the user to
simply press a single button to initiate the entire process. The button triggers
the image capture, OCR, and text-to-speech conversion automatically.
 HARDWARE & SOFTWARE REQUIREMENTS AND ESTIMATED
COSTS

HARDWARE

 Raspberry Pi 3 ₹3645
 Speaker ₹249
 Camera (OV5647) ₹200
 Cables and Connectors ₹100
 Adapter : ₹199
 Pvc box: ₹350
 Battery : ₹300
 Switch : ₹50

SOFTWARE

 Python 3 compiler
 Programming Language: Python
 METHODOLOGY &
IMPLEMENTATION
Block Diagram
1.Raspberry Pi 3B+
A single-board computer used as the core processing unit.
Features:
 Quad-core 64-bit processor.
 1 GB RAM.
 Multiple GPIO pins for interfacing with other devices.
 3.5mm audio jack for sound output.

2. Push Button
• Acts as an input device to trigger image capture or processing.
• Connected to the GPIO pins of the Raspberry Pi.
• Enables user interaction with the system.
2. Webcam
• Captures images as input for the project.
• Connected to the Raspberry Pi via a USB port.
• The images are processed by software running on the Raspberry Pi.

4. Speakers
• Outputs the speech generated from the image processing result.
• Connected to the Raspberry Pi through the 3.5mm audio jack or USB (if using a USB
speaker).
5. Power Bank
• Serves as the power source for the Raspberry Pi and its peripherals.
• Connected via the Raspberry Pi's power supply port (micro-USB or USB-C
depending on the model).

Connections in the Diagram:

• Power Supply: Power bank to Raspberry Pi for continuous operation.
• USB: For connecting the webcam to capture images.
• 3.5mm Audio Jack: For connecting the speakers to output audio.
• GPIO: Push button connected for user input.
FLOW CHART
 RESULTS & DISCUSSIONS
Experimental Setup:- The system was tested using text in Kannada, Hindi, and
English under various conditions to evaluate:
1. Text clarity and size.
2. Background complexity.
3. Lighting conditions.

Test Environment:
Lighting: Well-lit indoor settings with occasional tests in low-light conditions.
Text Source: Printed documents and posters in supported languages.
Camera: USB Webcam (720p resolution).

Parameters Evaluated:
OCR Accuracy: Ability to correctly extract text.
Language Detection: Ability to identify the dominant language..
TTS Clarity: Naturalness and intelligibility of spoken output.
RESULT ANALYSIS
Thresholds and Text Characteristics :-

Text Size: Text sizes between 12pt and 72pt (font sizes commonly used in printed
documents) were extracted with 95% accuracy.

Small Text Issues: Fonts smaller than 10pt resulted in OCR inaccuracies, with
recognition rates dropping 60% to 75%. Decorative or cursive fonts saw a drop in
accuracy to 70% due to OCR limitations.

Language Detection: Correctly identified the dominant language in 93% of cases

when single-language text was used. For mixed-language documents, results were
inconsistent, with ~75% accuracy in detecting the dominant language.

Translation: Kannada and Hindi translations to English showed ~90% accuracy for
simple sentences. Complex sentences with idiomatic expressions or ambiguous
contexts occasionally resulted in incorrect translations.
 FUTURE TRENDS

• Additional Language Support: Extend capabilities to support more regional and

global languages.

• Improved OCR: Integrate AI-powered OCR for better recognition of cursive or

decorative text and handwritten scripts.

• Compact Design: Develop an all-in-one device with integrated camera, speaker, and
tactile feedback for greater portability.

• Cloud Integration: Leverage cloud-based OCR and translation for faster processing
and real-time updates.

• Accessibility Features: Add voice commands, braille displays, or haptic feedback

for broader usability
 CONCLUSION

The Optical Image-to-Speech Converter Using OCR and TTS effectively transforms
printed text into audible speech, providing a valuable assistive tool for visually
impaired individuals.

The system demonstrated:

• Multilingual Support: Accurate recognition and translation of Kannada, Hindi, and
English text.
• High Performance: Reliable OCR and TTS outputs, with an average accuracy of
~90% for clear and well-lit images.
• User-Friendly Design: Interactive GPIO buttons and LEDs ensure ease of use and
portability.

By combining image preprocessing, language translation, and text-to-speech synthesis,

the project successfully created a portable, cost-effective solution for bridging
accessibility gaps in textual content.
 REFERENCES
 Ravi, S. Khasimbee, T. Asha, T. Joshna and P. Jyothirmai, "Raspberry pi based
smart reader for blindpeople," 2020 International Conference on Electronics and
Sustainable Communication Systems, no. 1, pp. 445-450, 2020.

 V. Mainkar, T. Bagayatkar, S. Shetye, H. Tamhankar andR. Jadhav, "Raspberry Pi

based intelligent reader for visually impaired persons," 2020.

 S. Akhil, "An overview of tesseract OCR engine," Department of CSE, Calicut

Monsoon, 2016.

 S. Anbarasi, R. Krishnaveni and R. Aruna, "Smart Reader Glass for Blind and
Visually Impaired People," IOS Press, 2021.

 http://www.daveconroy.com/turn-raspberry-pi-translator-speech

 www.raspberrypi.org
 O U
K Y
A N
T H