International Conference on Electrical, Electronics, Signals, Communication and Optimization (EESCO) - 2015
978-1-4799-7678-2/15/$31.00 ©2015 IEEE
Multilingual Speech Translator using MATLAB
Rubi Debnath, Vivek Hanumante, Disha Bhattacharjee, Deepti Tripathi, Sahadev Roy
Department of Electronics & Communication Engineering
National Institute of Technology, Arunachal Pradesh
Yupia, INDIA
Abstract—We design a translator using MATLAB to
overcome the barrier of different languages. The proposed
design may be use for English text to speech conversion into
multiple languages. User able to input English text manually to
the GUI or image of printed text as input and output as
translated speech in different language by help of Optical
Character Recognition (OCR) technique. The proposed
methodology may be use to provide assistance to people who
lack the power of speech or non-native speakers.
Keywords—GUI; MATLAB; T2MST; Multilingual; OCR;
Text To Speech; sample frequency
I. INTRODUCTION
English being the most widely spread language has been
the center of attraction for synthesizes speech, these system
is most commonly known as Text to Speech converter [1,
2]. There are plenty of options which are able to translate
text into other languages [3-6]. Majority of text to speech
converter suffer from the limitation of being able to convert
the text to speech in to a particular language. On the other
hand some available multilingual translators are able to
translate only text to text or speech to speech only [7, 8].
Since each type of system uses different sets of algorithms,
these algorithms not employed to translate languages along
with text to speech converter.
GUI of Text to Multilingual Translator (T2MST) has
both facilities to translate from manually input English text
or image of printed text and provide output in multiple
languages in the form of speech. In this way we are
facilitating language translation and text to speech
conversion simultaneously. By the proposed algorithm
T2MST will be able to recognized manually enter the text or
from which text can be extracted using Optical Character
Recognition [9]. The system presently is able to provide
speech output in three different languages, namely –
English, Hindi and Bengali, three most spoken languages in
India. User can select the language for which he wishes to
hear output voice.
The system can be further extended to include more
languages and dialects to increase the usability of system.
Further developments in the system requires implementation
of grammar rules and generation of natural voice [10].
II. OVERVIEW OF T2MST
The proposed T2MST is a hybrid system of text to
speech conversion with translators. This system has been
designed using MATLAB. Architecture of the proposed
design is as fig 1. Currently developed sound library
includes most commonly used words in daily life
conversations. User is asked to select a language first out of
the provided options. User can enter the text manually or
can choose to read text from an image of printed text. Based
on user selected method system acts further and provide
speech output translated based on the language selected by
user.
Fig.1. System Architecture of Text to Multilingual Speech
Translator
III. DESIGN METHODOLOGY
The proposed methodology of the system deploys two
methods for speech translation of input text. These two are
based on stored sound library and MATLAB TTS function
[11]. Firstly the translation is done on the basis of iterative
comparison with the words stored in the sound libraries. If
word is not found in the stored in library then the word is
pronounced using syllable based method of MATLAB TTS
function. For example the words which are the names of a
person or place etc. which are most probable not to be found