Browsing by Author "Ariyarathna, G.D.W.M."

Now showing 1 - 4 of 4
  • No Thumbnail Available
    Item
    Automated Cephalometric Analysis in Orthodontics Using Artificial Neural Network
    (Uva Wellassa University of Sri Lanka, 2010) Ariyarathna, G.D.W.M.; Karunananda, A.S.
    This work demonstrates the use of Neural Network approach, which is being developed to promote the automated identification and localization of cephelametric landmarks in orthodontics. Orthodontics is a specialty of dentistry that is concerned with the study and treatment of malocclusions (improper bites), which may be a result of tooth irregularity, disproportionate jaw improved bite (occlusion). Identification and localization of cephelametric landmarks has become an important clinical task in orthodontics. The conventional method of locating landmarks depends on manual tracing of the radiographic images. Since this is time consuming and error proven, the demand for completely automated analysis and diagnostic tasks has increased. In this respect, an intelligent cephalometric analysis is one of the main goals to be reached in orthodontics in near future. This work critically reviews four major problems in cephelametric analysis namely; precision of landmark identification and localization, enormous time consumption, human errors and need for continuous support from experts. We argue that, issue of lack of autonomated solutions for cephelametric analysis has been the main problem.Conventional approaches lack generality, adaptability and flexibility, since it is difficult for them to learn the environment changes and they do not provide facilities to automate the analysis process to improve the accuracy of landmark detection. There have been several previous attempts to automate cephalometric analysis with the use of hand crafted algorithms, mathematical or mtatistical models and artificial intelligence techniques such as neural networks, genetic algorithms and fuzzy logic. Nevertheless, in all cases accuracy was the same or worse than that of manual identification. Therefore this investigation aimed at proposing an Artificial Neural Network approach to computerize the cephalometric analysis. It is evident from the literature that, neural network approach can introduce very high level of autonomy and accuracy in modeling real world problems. The proposed system automates cephalometric analysis along four dimensions; image acquisition: capturing and scanning cephalograms, image processing and computer vision: image analyzing, edge detecting and extracting landmarks, ANN training: classifying landmarks according to their geometrical specifications and pinpointing the land marks by calculating the center of gravity in each cluster. The users of the system would be orthodontists. This system has been implemented as a desktop application which automatically analyzes the cephelametric land marks according to their geometrical classifications. Key words: Automated cephalometric analysis, Artificial neural network
  • No Thumbnail Available
    Item
    Forecasting Stock Market Indices Using Artificial Neural Networks
    (Uva Wellassa University of Sri Lanka, 2010) Madhuranga, G.G.A.G.; Ariyarathna, G.D.W.M.
    This research work presents an Artificial Neural Network (ANN) approach for stock price indices forecasting. The data from the Colombo stock exchange (CSE) have been used as inputs for the Proposed System. The nonlinear nature and the complex behavior of this type of time series data, create difficulties in forecasting. This research determines the feasibility and practicality of using an ANN as a forecasting tool since it is capable of approximating nonlinear functions. Aim of this investigation was to demonstrate the utility of Feed Forward Back Propagation neural network to investigate the predictability of stock market indices using the most appropriate ANN model that would be able to accurately predict the future behavior of stock market price indices. This has been accomplished by developing a system for time series forecasting with the use of neural network technology. The system was implemented by considering main properties of ANN such as choice of input parameters, determination of the number of neurons in hidden layer, learning rate of neural network training and finally construction of a neural network forecaster. Large amount of neural networks were implemented with different combination of properties as mentioned above and finally the most optimum model was selected as forecaster. Implemented system successfully achieves the good directional prediction accuracies. Computer simulations are presented to show the effectiveness of the implemented system. This study shows that the accuracy of the model mainly depends on the input parameters and the architecture of the neural network. Finally it suggests ways to improve the ANN forecasting model. Key words: Artificial Neural Network (ANN), Colombo stock exchange
  • No Thumbnail Available
    Item
    University Timetabling System using Genetic Algorithm
    (Uva Wellassa University of Sri Lanka, 2010) Welenawewa, N.I.K.; Ariyarathna, G.D.W.M.
    Setting up the timetable has been a real burden to the lecturers and a distraction to lecturers' core responsibility of teaching in almost all universities of the country. This project, "University Timetabling System using Genetic Algorithm" aims to review the current manual timetabling system of the Division of Computer Science & Technology (DSCT) and develop a web based timetabling system using genetic algorithm. Data gathering for the system was done through interviewing relevant stakeholders and from the literature review. Gathered data were analyzed as input to the proposed DCST system. The literature review was carried out to search the best approach that can help to solve the problem in the timetabling system including the heuristic approach, integer programming approach, graph coloring approach, network streaming approach, logical constraints arithmetic approach, knowledge based approach, tabu searching, annealing simulation and genetic algorithm. It is evident from the literature that, Genetic Algorithm can introduce very high level of autonomy and accuracy to produce a feasible timetabling system. The improved DCST timetabling system was implemented using PHP programming and Visual Studio C++. Three modules have been developed; namely: Administrator Module, Lecturer Module and Student Module. The lecturer module will set the master timetable for all the lectures in the DCST. Similarly through the student module, students can view their own timetable for the whole semester. In the Administrator Module, the administrator can manage the student registration, lecture and subject registration by adding or deleting lectures or subjects. Genetic Algorithm has been used due to the ability of producing optimum solution to generate a feasible timetabling system. It is hoped that the proposed timetabling system will help lecturers to concentrate in their core activities of teaching and research rather than spending more time on administrative work such as preparing timetables. Key words: PHP Programming, visual studio C++, genetic algorithm, timetabling
  • No Thumbnail Available
    Item
    User Authentication via Speaker Identification
    (Uva Wellassa University of Sri Lanka, 2010) Aththanagoda, A.K.N.L.; Ariyarathna, G.D.W.M.
    Speech is the primary mode of communication among humans, and voice modalities seem to be the most convenient for the users in such authentication system. Therefore, the concept of automatic speech identification was developing rapidly in last few decades. The automatic speaker recognition technologies have become more important and speech aided applications are being used for many researches. The main challenge of automatic speaker recognition is to deal with the variability of the environments and channels from where the speech is obtained. This research presents speaker recognition system that has been developed by using artificial neural network. It consists of three main modules namely LPC module, neural network module and final GUI module. The final user identification module presents the output generated by the other modules. When considering the first two modules, module two recognizes speaker using Back- propagation Neural Networks (BNN) in which input signals are out coming from a Linear Predictive Coding (LPC) processing method (module one) that characterizes each voice signal. However, these three modules operate separately. One of the important specialty of this system is all three modules were developed using MATLAB programming language. Even for the user interfaces this system has used MATLAB. The reason is we need to prove that, when developing an ANN system it should not depend on the UI developing with another programming language. The speaker identification accuracy is around 85% for the current developed system. Still this system has limitations The number of users that can use this system is still limited. The user can add new users to the system by using all three modules in order. The range of accept and reject is defined by using the simulation results of the neural network. If the numbers of users increase, the similarity will increase, and cause limitation in this system. Key words: MATLAB programming language, Back- propagation Neural Networks (BNN), Linear Predictive Coding (LPC) processing