Centre of Excellence in Cognitive Science - CECoS

College of Research Methodology and Cognitive Science (RMCS), Burapha University

 

  Home   Event    People    Research   Project  ○  Publication  ○  Collaboration  ○  Contact Us

 

Senior engineering projects done in 2010

Senior engineering projects done in 2009

Senior engineering projects done in 2008

Copyright © All rights reserved.

 

" Robust Thai Speech Recognition System against Real Environmental Noise "

Project team: Sukanya Taepaiboon, Phraewpran Khamjaroen, and Chanita Khemsamrit

URL: http://www.rmcs.buu.ac.th/cecos/project/tsp (Thai)
Demonstration video (Thai)


 

Problem: Intelligent technology in our daily life is growing rapidly, including an automatic speech recognition (ASR) systems in which we can apply ASR for many electronic devices around us, for example, speech-to-text in computers, voice-controlled dialing in mobile phones, and voice-controlled commanding in robots. As for the related works, the system’s performance of accuracy rate in classifying speech signal will be attenuated when the background noises are present.

Basic Idea: This project applied the traditional enhancement on the frequency domain, namely the spectral subtraction (SS) method. By applying this method, we used a two-channel ASR based system. Due to the noiseless and noisy speeches are recorded separately as depicted in figure below, with this concept idea the noisy speech can be enhanced by means of the SS method.

System overview of this project

Method: The cepstrem parameter of enhanced speech was used as a feature then classify these features by using minimum sum of squared difference (SSD). Speech signal was digitized at a sampling rate of 11,025 Hz, 16-bit resolution, with a total length of processing of one second. We designed the software program “Genius Speech”.

SS method originally proposed by Boll (IEEE Trans. on ASSP, 1979)

 

Training page

 

Processing steps of software: 1) Training, 2) Enhancement, 3) Classifying, 4) Output result
 

Experimental Results: In the experiment, 10 males and 10 females volunteered to be subjects. Thai words, “0” to “10”, “Open”, and “Close” were used as input speech data. The experimental results show that the proposed ASR system can achieve average accuracy rates of 96.92% in noiseless environment, 82.71% in air condition noise, 79.83% in babble noise, 73.75% in construction area noise, and 72.25% in traffic noise.

Publication: M. Phothisonothai and C. Fongsamut, "Robust Thai Speech Recognition Method against Real Environmental Noises", The 6th National Conference on Computing and Information Technology (NCCIT), vol.1, pp.205-210, Bangkok, Thailand, 2010.
 

 

" Development of 2D Surface Image Analyzing Method Based on Fractal Theory "

Project team: Saranya Winijmongkolsin and Supichaya Areewongsin

URL: Demonstration video (Thai)


 

Problem: Digital image processing technology plays an important role in many fields, for examples; the inspection process of mammogram images when  cancer is found in medicine, the classification process of optical character recognition in computer science, and the inspection process of the quality of fruits in agriculture.Also, featuring process in image analyzing is an important step especially surface image analyzing method based applications.

Project's framework

Basic Idea: This project, therefore, proposes a method to analyze a 2D surface image or texture on the basis of fractal geometry concept in which the corresponding dimensional value, called fractal dimension (FD), has been determined. The FD has proven that the complexity of surface image can be quantified effectively: also can improve the output feature of a surface in terms of self-similarity, increasing reliability.

 

FD evaluation of image captured by web camera (Online processing)
 

FD evaluation of image in directory file (Offline processing)
 

Method: In the experiment, the input image will be captured via web camera OKER model OE-184 with 24-bit color and a resolution of 480x640 pixels. Due to the non-stable factor of the image acquisition process, for example, intensity of light can directly affect the quality of the input image. Thus, the captured image will be processed by pre-filtering, shading correction, and histogram based on a gray-to-binary conversion, respectively. The FD in this project has been estimated by the box-counting method (BCM).

Experimental results, images were captured by web camera

Type of input image True FD % Error
Koch snowflake curve 1.260 3.462
Sierpinski triangle 1.585 3.546
Koch curve 1.500 1.507
Apollony 1.3058 1.130
Apollonian gasket 1.400 0.623


Experimental Results:
Based on the experimental results, when compared we found an average error rate of approximately about 3.785% and 3.100% for comparison with the original image files and the image captured by web camera, respectively. Moreover, to show the effectiveness of the proposed method in real applications, we have investigated the FD values of the images of characters in different languages.

 

" The Design of Temperature Monitoring System for Vaccine Storage Refrigerator "

Project team: Atidtaya Jitpirom, Surang Saranpong, and Chawanwit Panduangnet

URL: http://www.rmcs.buu.ac.th/cecos/project/factory (Thai)


Abstract:
A vaccine is a biological preparation that inoculates humans or animals to a particular disease. To maintain the quality of a vaccine, a temperature monitoring process is very importance to get the maximum benefit of treatment. Therefore, this project proposes a simple design of temperature management and monitoring system for the refrigerator that storages vaccines. Six temperature sensors were installed for each refrigerator in which the interfacing process of signal was controlled by microcontroller. All temperature data were formed into a single package containing temperature data, and the specific characters to notify the beginning and the end separate packets. This data will be sent via RS232 circuit and then converting by RS485 circuit. The acquired temperature will be shown as a line graph, and will be saved to the database in a control unit (computer). If the observed temperature of the refrigerator was higher than the critical temperature continuously over 20 minutes then program will send SMS automatically to the administrator’s mobile via a GSM Module. In this system, users are able to retrieve the historical data, set the critical temperature for each vaccine storage refrigerator, and set the mobile phone number for sending SMS. The system has been installed at Chonburi hospital.

 

 

LPN Software designed by our team

 

Copyright © 2009-11 CECoS, College of Research Methodology and Cognitive Science, Burapha University, Chonburi 20131 Thailand.