International Journal of Advances in Electronics and Computer Science, ISSN: 2393-2835 Volume-4, Issue-4, Aprl.-2017 http://iraj.in A TALKING DICTIONARY FOR ENGINEERING TERMS

1MD. ABDULLAH AL NOMAN, 2RUBEL KANTI DAS, 3A.S.M NOMAN, 4MITUN SHIL

1, 2, 3, 4Southern University Bangladesh, Chittagong, Bangladesh Email: [email protected], 2 [email protected], 3 [email protected], [email protected].

Abstract - This paper presents a realistic translator of different engineering terms. The device can understand the users need and if the data is present in its database then it gives the explanation of the given task. Moreover, if the answer that needs to be replied is not in the database, then the dictionary asks the user if the user wants to record the answer. Thus the database of the dictionary can be enriched. has been used as core controller. The system consists of STT and TTS.

Indexterms - Dictionary, Talking dictionary, Raspberry Pi, STT, TTS.

I. INTRODUCTION To Text) system [6]. On the contrary, python-pymad has been installed for TTS (Text To Speech) system In enpgineering discipline we need to learn different [7]. engineering term. In that case only dictionary book is one of themost important way. Utility conversional III. OVERVIEW OF THE SYSTEM text-based dictionaries could be greatly enhanced by means of electronic dictionary using speech as mode Basically the device is a talking dictionary which can of interaction. This paper presents a talking dictionary interact with the user. If any question is asked and if as an alternative to conventional dictionary texts. the answer of that question is in the database then it reads out the answer with a realistic voice along with In this paper a talking dictionary is presented that can understandable speed. Besides if the answer of the understand the question asked by user and give the question that has been asked is not in the database, answer by talking with realistic voice [1, 2]. then the device asks the user whether the user wants to record the answer of that question or not. If the The paper is described in four sections where section user wants to record the answer then the device II delineates about the system, section III illustrates records the answer and store it in the database. regarding the working procedure and section IV concludes the study. IV. WORKING PROCEDURE

II. LITURATURE REVIEW Raspberry Pi [8-10] has been employed as core controller of the device. It supports a dozen of based embedded platform is used in this programming languages like C, C++, JAVA, PHP project. Embedded Linux uses the Linux kernel and and many more [11]. various open-source components in embedded systems [3]. When the device is activated it initializes the I/O and looks for if there is any error or not. If the is an error It is an assembled under the model then it initializes the I/O again. If there is no error of open source software development and mostly then it starts listening from the users. But the device used for server purpose. There are Different types of does not accept the question unless it gets the pass Embedded Linux systems like ETLinux, LEM, key. The user has to mention pass key first and then LOAF, uClinux, U Linux, Thin Linux. Some has to ask the question. If the pass key is pronounced, preferred platforms are: Google's Android (operating the device starts STT (Speech To Text) [12]. system) well-known type of embedded Linux, e.g. on , Tizen – embedded Linux for Then the device does spectral analysis. According to smartphones, – used on Raspberry Pi, the spectral analysis the STT (as query) is searched in Emdebian Grip, BusyBox, , Familiar the database. If the query is present in the database Linux, Mobilinux[4]. then it replies by reading out the answerusing TTS (Text To Speech) system. The query is searched Some widely used Linux Based Embedded Devices based on the spectral analysis. If the spectrum meets are: Arduino Yun, BeagleBoard-xM, BeagleBone A6, the accuracy around 80% with the spectrum of the Cubieboard, pcDuino, RascalMicro, Raspberry Pi database, then it selects that particular spectrum from [5].Besides using Linux, speech recognition can be the database as its preferred answer. easier. In this case we have employed raspberry pi using Linux operating system. Besides ‘pocketsphinx’ library has been used for STT(Speech

A Talking Dictionary for Engineering Terms

15 International Journal of Advances in Electronics and Computer Science, ISSN: 2393-2835 Volume-4, Issue-4, Aprl.-2017 http://iraj.in

Figure 1: working system of the device Figure 2: Algorithm for the system

In figure 1, it has been shown that how the system V. APPLICATION DOMAIN works in a nutshell. For the sound processing a signal processing algorithm is used. After processing the It is known that interactive learning is more sound in STT, the string is given a particular score understandable and can be memorized for long time. known as acoustic scoring for searching in the The visualization method is making more sense in database. In this case the ‘spans on’ technique is learning. As studying from dictionary book is not used. For the scoring a scoring algorithm is used and interesting all the timein that case the talking an acoustic database is used to find the matches. For dictionary can be a great source of learning with fun searching a searching algorithm is used and the due to having interaction system with the user and it database used is language and dictionary database. is very user friendly. It works both in offline and online mode.

VI. RESULTS AND DISCUSSION

The device has been tested using two users to test the accuracy of given task which has been shown in the table 1, table 2& table 3. Accepted & non-accepted rate are shown in there as well. Table 4shows the percentage comparison of the spoken acceptance rate. The overall acceptance rate obtained is 80%.

Table 1 Accuracy testing for spoken terms by trial 1

TRIAL NO:- 01 FREQUENCY SPOKEN TERM NOT- NO ACCEPTED ACCEPTED ELECTROMAGNETIC 10 9 1 GALVANOMETER 10 8 2 EARTH GROUND 10 7 3 ECHO 10 10 0 VOLTAGE 10 8 2 Figure 3: Prototype

A Talking Dictionary for Engineering Terms

16 International Journal of Advances in Electronics and Computer Science, ISSN: 2393-2835 Volume-4, Issue-4, Aprl.-2017 http://iraj.in

Table 2 CONCLUSION Accuracy testing for spoken terms by trial 2

In this paper a talking dictionary is described which TRIAL NO: - 02 has the interaction facility with the user. Besides it FREQUENCY can be handled very easily. Further we are working to add more intelligence like understanding different SPOKEN TERM NOT- ways of asking the same question, understanding and NO ACCEPTED ACCEPTED talking in different language. ELECTROMAGNETIC 10 8 2 GALVANOMETER 10 8 2 REFERENCES EARTH GROUND 10 6 4 ECHO 10 10 0 [1] S.Suresh, Y. Sindhuja Rao, “Modelling Of Secured Voice Recognition Based Automatic Control System”, IJETCSE, VOLTAGE 10 8 2 ISSN: 0976-1353, Volume 13, Issue: 2 –MARCH 2015 [2] Anusha Bhargava, Karthik V. Nath, Pritish Sachdeva and Table 3 Monil Samel, “Reading Assistant for the Visually Impaired”, Accuracy testing for spoken terms by trial 3 IJCET, E-ISSN 2277 – 4106, P-ISSN 2347 – 5161, Vol.5, No:- 2, April 2015 TRIAL NO: - 03 [3] PramodKumar Singh, Prof SwaniliKarmore, “Startup Time FREQUENCY Optimization Techniques for Embedded Linux” ,

SPOKEN TERM International Journal of Scientific & Engineering Research, NOT- NO ACCEPTED ISSN 2229-5518, Volume 7, Issue 3, March-2016 ACCEPTED [4] Embedded system, Wikipedia, USA, 15.01.2017, https://en.wikipedia.org/wiki/Linux_on_embedded_systems ELECTROMAGNETIC 10 8 2 [5] Linux Embedded Devices Comparison: Yun, BeagleBoard, Rascal, Raspi, Cubieboard&pcDuino, cooking hacks, GALVANOMETER 10 8 2 Zaragoza (Spain), 18.01.2017, https://www.cooking-hacks.com/blog/new-linux-embedded- EARTH GROUND 10 6 4 devices-comparison-arduino--rascal-raspberry-pi- cubieboard-and-pcduino/ ECHO 10 9 1 [6] PrachiKhilari , Prof. Bhope V. P., “Implementation of Speech to Text Conversion” , IJIRSET, ISSN(Online) : 2319-8753, ISSN (Print) : 2347-6710, Vol. 4, Issue 7, July 2015 VOLTAGE 10 8 2 [7] Suhas R. Mache, Manasi R. Baheti, C. NamrataMahender, “Review on Text-To-Speech Synthesizer” , IJARCCE, ISSN Table 4 (Online) 2278-1021, ISSN (Print) 2319 5940, Vol. 4, Issue 8, Accuracy (in %) for spoken term August 2015. [8] Richard Grimmett, “Raspberry Pi Robotics Project”, second edition, 2015, ISBN: 978-1-78528-014-6 [9] Aamir Lakhani, Joseph Muniz, “Penetration Testing with TRIAL TRIAL TRIAL SPOKEN TERM Raspberry Pi”, ISBN: 987-1-78439-643-5, January 2015 NO: 01 NO: 02 NO: 03 [10] Peter Membrey, David Hows,“Learn Raspberry Pi with Linux”, ISBN: 978-1-4302-4821-7 ELECTROMAGNETIC 90% 80% 80% [11] What to choose linux operating system or windows operating GALVANOMETER 80% 80% 80% system? , Host Department, USA, 16.01.2017, EARTH GROUND 70% 60% 60% [12] https://www.hostdepartment.com/blog/2013/12/27/choose- ECHO 100% 100% 90% linux-operating-system-windows-operating-system/ VOLTAGE 80% 80% 80% [13] DeepaV.Jose, Alfateh Mustafa, Sharan R, “A Novel Model OVERALL for Speech to Text Conversion” ,IRJES, ISSN (Online) 2319- 84% 80% 78% PERCENTAGE 183X, (Print) 2319-1821, Volume 3, Issue 1 (January 2014), PP. 39-41



A Talking Dictionary for Engineering Terms

17