Conference Papers

Permanent URI for this collection

Browse

Recent Submissions

Now showing 1 - 20 of 24
  • Item
    Potential Z-Fighting Conflict Detection System in 3D Level Design Tools
    (2015) Pisal Setthawong
    Z-Fighting is an effect that happens in 3D scenes when two co-planar surfaces share similar values in the z-buffer which leads to flicking and visual artifacts during the rendering process due to conflicting order of rendering the surface. However in 3D level design, scenes created by the tools can be complex, in which level designers can inadvertently place co-planar surfaces that would be susceptible to z-fighting. Level designers typically notice the z-fighting artifact through visual inspection through the usage of a 3D walkthrough test on the scene which is time-consuming and easy to miss. To solve the issue, a proposal of a zfighting detection system for level design tools is proposed to streamline the process of detecting potential hotspots where z-fighting conflicts may occur from co-planar objects.
  • Item
    Optimizing Player Throughput for Interactive Motion Based Kiosk Games – A Case Study from the PTT Technobots Campaign
    (2015) Pisal Setthawong
    Maintaining booths that are attractive to exhibition attendees is one of the main goals of exhibitors. One of the popular tactics to attract attention is to utilize new and emerging technology to create fresh new types of interactive booths, in which motion-based kiosk games are gaining popularity in this domain. However motion-based kiosk games are usually designed using conventional computer game design principles and are not optimized for exhibitions which results in a low player throughput. This paper examines the issues behind conventional game design principles, and proposes changes that would improve player throughput and is empirically validated upon a real world case study in which the author has worked on. The case study selected is the Petroleum Authority of Thailand (PTT) Technobots Campaign, in which a series of motion-based kiosk games were deployed at a number of popular department stores in Bangkok, Thailand during the period of May-Aug 2013.
  • Item
    Domain-oriented two-stage aggregation: generating baseball play-by-play narratives
    (2015) Baldwin, James; Songsak Channarukul
    This paper presents an end-to-end natural language generation system that performs aggregation in two stages: the first takes advantage of the information implicit in the source knowledge base in order to aggregate event components into complex sentences. The second stage examines the developing context of the text in order to aggregate similar adjacent events into more fluent text. The source knowledge base is the Retrosheet collection of play-by-play baseball scoresheets encoded in machine-readable form. The output is reasonably fluent and natural, human-readable play-by-play narratives of historical baseball games. The system was tested against all regular season major league games played from 1950 to 1969, taking less than a second to produce three to five pages of text for each game. The aggregation achieved resulted in a substantial improvement in native speaker judgments of fluency and readability.
  • Item
    Alternating Least Squares with Incremental Learning Bias
    (2015) Aung, Than Htike; Rachsuda Jiamthapthaksin
    Recommender systems provide personalized suggestions for every individual user in the system. Many recommender systems use collaborative filtering approach in which the system collects and analyzes users' past behaviors, activities or preferences to produce high quality recommendations for the users. Among various collaborative recommendation techniques, model-based approaches are more scalable than memory-based approaches for large scale data sets in spite of large offline computation and difficulty to update the model in real time. In this paper, we introduce Alternating Least Squares with Incremental Learning Bias (ALS++) algorithm to improve over existing matrix factorization algorithms. These learning biases are treated as additional dimensions in our algorithm rather than as additional weights. As the learning process begins after regularized matrix factorization, the algorithm can update incrementally over the preference changes of the data set in constant time without rebuilding the new model again. We set up two different experiments using three different data sets to measure the performance of our new algorithm.
  • Item
    Feature Envy Factor A Metric for Automatic Feature Envy Detection
    (2015) Kwankamol Nongpong
    As a software system evolves, its design get deteriorated and the system becomes difficult to maintain. In order to improve such an internal quality, the system must be restructured without affecting its external behavior. The process involves detecting the design flaws (or code smells) and applying appropriate refactorings that could help remove such flaws. One of the design flaws in many object-oriented systems is placing members in the wrong class. This code smell is called Feature Envy and it is a sign of inappropriate coupling and cohesion. This work proposes a metric to detect Feature Envy code smell that can be removed by relocating the method. Our evaluation shows promising results as the overall system’s complexity is reduced after suggested Move Method refactorings are applied.
  • Item
    The Effect of Transmission Range in Multi-hop Wireless Networks
    (2014) Kunagorn Kunavut
    Transmission or communication range is an important factor for successful data delivery in wireless communications including multi-hop wireless networks. Typically, transmission range can be technically adjustable by configuring the transmitting power (Tx Power) or changing the antenna height. For the same antenna height, if transmission range is minimized or adjusted to be shorter by lowering Tx power, there is less energy consumption but the networks are likely to be unconnected which consequently degrades the network performance such as throughput and delivery ratio. In the case that range of nodes is maximized or extended for connectivity by increasing Tx power, networks become connected and a node may reach the others by using just a hop. This beneficially affects network performance but these nodes with range extension consume more energy for data transmission. Hence, there is trade-off between energy consumption and network performance when adjusting transmission range. In addition, in multi-hop wireless networks where nodes are usually mobile and network topology are highly dynamic, transmission range has lots of impact on network performance. Hence, to study the effect of transmission range is very important and required. In this work, various scenarios (i.e. load-, speed- and density-varying scenarios) are constructed to investigate both energy consumption and network performance with different transmission ranges.
  • Item
    Performance and Comparative Analysis of Energy Consumption for Mobile Ad Hoc Routing Protocols
    (2014) Kunagorn Kunavut
    Mobile ad hoc networks are energy-constrained networks, nodes in these networks are usually mobile and they are typically implemented in the specific areas where electrical power source is not available. Thus, the operations of these nodes must only rely on battery power or other exhaustible sources of energy. In addition, most of power in each node is required for communication with other nodes or gateway connected to the wired networks, this usually includes transmission of data and control overhead generated by ad hoc routing protocols to obtain the best paths to destination nodes/devices. Thus, to study the energy consumption of each routing protocol is one of the major concerns. In this work, number of simulation scenarios are constructed to study the energy consumption in each node using various routing protocols standardized by the Internet Engineering Task Force (IETF) MANET working Group which are OLSR (RFC 3626), DSR (RFC 4728) and AODV (RFC 3561).
  • Item
    Performance Evaluation of Dynamic Cell Zooming Algorithms in Omni-directional and Sector-based Cells
    (2014) Tun, Khin Cho; Kunagorn Kunavut
    This paper evaluates and highlights the performance of three dynamic cell zooming algorithms applied in both omni-directional and sector-based networks. A possible framework compatible with dynamic cell zooming algorithms for user’s location detection is presented. The performance of each cell zooming algorithm is simulated in terms of power saving and possible outage in a full-day operation. According to simulated results, there is no significant difference between the performance of each algorithm and others at low traffic hours, but their performances are different at high traffic hours. From an overall comparison, the continuous cell zooming algorithm illustrates the best performance in terms of power saving, followed by fuzzy algorithm and then discrete algorithm. However, in terms of possible outage, the continuous algorithm is very sensitive to user movement and it shows a very high possible outage ratio. Meanwhile, the outage is totally removed in the discrete and fuzzy algorithms according to their concept. The dynamic cell zooming algorithms show a larger power saving in sector-based network since it hosts a more detailed plan to perform cell zooming.
  • Item
    Adaptive Location Update Schemes for Continuous Cell Zooming Algorithm in Wireless Networks
    (2015) Tun, Khin Cho; Kunagorn Kunavut
    Continuous cell zooming algorithm is a potential dynamic cell zooming algorithm for energy-efficient operation of mobile wireless networks. In this algorithm, location management strategy (location update process) is required to know the location of the farthest user in each cell to perform cell zooming. However, the application of conventional periodic update scheme in continuous cell zooming algorithm can lead to a high signaling cost. Therefore, in this paper, two adaptive location update schemes, namely, Time-Adaptive Periodic Update (TAPU) and Location-Adaptive Periodic Update (LAPU) are proposed aiming to reduce the number of update messages in continuous cell zooming operation. The performances of the proposed adaptive location update schemes are compared with that of Convention Periodic Update (CPU) scheme. Their performances are evaluated in terms of power saving capability, outage ratio and number of update messages raised in cell zooming operation in both omni-directional and sector-based cell networks. The TAPU and LAPU have no significant effect on power saving capability of continuous cell zooming algorithm and they give less number of update messages compared to CPU scheme. However, outage occurs in cell zooming operation with TAPU scheme because it has longer update intervals. Meanwhile the LAPU scheme can eliminate outage in cell zooming operation as CPU scheme does.
  • Item
    License Plate Extraction and Recognition of a Thai Vehicle Based on MSER and BPNN
    (2015) Tao, Hong; Gopalakrishnan, Anilkumar Kothalil
    The extraction and recognition of a Thai vehicle license plate based on Maximally Stable Extremal Regions (MSER) and Back-Propagation Neural Network (BPNN) is presented. The license plate area is in a maximally stable extremal region of a car image. It can be effectively extracted from MSERs by multiple classifications. The feature extraction of characters from the license plate is based on Zernike moment. The feature is used as a training dataset for the BPNN to recognize the characters. The experimental results indicate that the proposed approach is an effective method for the extraction and recognition of a Thai license plate.
  • Item
    A Collaborative Filtering Recommendation Based on User Profile and User Behavior in Online Social Networks
    (2014) Lu, Yang; Gopalakrishnan, Anilkumar Kothalil
    This paper aims to present and discuss the similarity among users in a social network based on CF (Collaborative Filtering) algorithm and SimRank (Similarity Based on Random Walk) algorithm. The CF algorithm used to predict the relationship between users based on the user rating on items (movies and books) and the user’s profile. The SimRank algorithm calculates the similarity among users through finding the nearest neighbors for each user in the social network. At last, the combination of these two algorithms will be used to get “people may interest each other” from users’ database. In the experimental analysis, a data set “DouBan” (a data set is collected from a Chinese website) will be used and demonstrates the performance of the improved technique with a website. And the website will be developed to show the recommended processing of the proposed algorithm. Finally, the recommendation accuracy of the proposed method is evaluated by comparing with the existing recommendation algorithms.
  • Item
    Enhanced Fuzzy-based Handover Decision System Design for Wireless Mobile Networks
    (2015) Thanachai Thumthawatworn; Piyakul Tillapart; Pratit Santiprabhob
    Numerous fuzzy-based algorithms have been suggested to enhance the intelligence of handover decision system for wireless mobile networks. However, most existing algorithms are based on a monolithic fuzzy engine design resulting in a large algorithm execution time, if a number of decision parameters is large and gives a poorer network selection performance, when dealing with different traffic types. In this paper, an enhanced fuzzy-based handover decision system design is proposed to address the above issues. The proposed design is presented and the simulation results show that the proposed design enhances the network selection performance and reduces the algorithm execution time.
  • Item
    Intelligent Handover Decision Based on Fuzzy Logic for Heterogeneous Wireless Networks
    (2015) Piyakul Tillapart; Thanachai Thumthawatworn; Piboonlit Viriyaphol; Pratit Santiprabhob
    An intelligent hand over decision system (HDS) is essential to heterogeneous wireless mobile networks in order to fulfill user's expectations in terms of universal and seamless services. With emerging real-time services in heterogeneous networking environment, including multiple QoS parameters in hand over decision process seems essential. In this paper, fuzzy logic is applied to enhance the intelligence of HDS. A new fuzzybased HDS design with the aim to reduce design complexity of fuzzy engine without sacrificing handover decision performance is proposed in the paper. The results show that, compared to non-fuzzy-based (i.e., SAW and AHP) and existing fuzzybased decision techniques, the network selection performance of proposed HDS design is significantly better than SAW and AHP, and is superior to an existing fuzzy-based technique. The proposed HDS design is then enhanced by incorporating an adaptive mechanism enabling a further improvement in terms of the network selection performance.
  • Item
    Performance Evaluation of Adaptive Rate Control over Uncompressed High-Definition Content Transmission with Paralleled Digital Subscriber Lines
    (2015) Surasee Prahmkaew; Siriwhaddhanah Pongpadpinit; Piboonlit Viriyaphol; Chanintorn Jittawiriyanukoon
    High-Definition (HD) information is the consumed bandwidth application in Telecommunication network. The quality of service (QoS) on uncompressed High-Definition content transmission with Paralleled Digital Subscriber Lines is needed in smooth information transmission and most tasks on QoS are served by each telecommunication network operator. The Adaptive Rate Control (ARC) is introduced and deployed on High-Definition (HD) contents transmitting on Paralleled Digital Subscriber Lines and compare with classical mechanism to provide more reliability, utilization, and throughput before served in telecommunication network. In this paper we simulated the application of the paralleled Digital Subscriber Lines (DSLs) to deliver the HD information, As results, the performances of communication are evaluated base on throughputs, utilization, and queue performance matrices. Finally results were impressive to support by ARC.
  • Item
    A Prediction System Based on the TMRS Standardized Socio-Economic Status (SES) Classification of Bangkok and Metropolian Subjects
    (2015-01) Jirayut Poomontre; Pisal Setthawong
    The paper proposes a prediction system for socio-economic status (SES) classification of Bangkok and metropolitan subjects. The SES classification is based on the standardized SES classification that was proposed by the Thailand Marketing Research Society (TMRS) with the support of the National Statistical Office (NSO), and widely adopted by local marketing research firms. Extending the author’s previous work on the standardized TMRS SES classification, the paper describes a prediction system that was developed to classify Bangkok and metropolitan subjects into the SES classes that was proposed earlier.
  • Item
    Robust Block-Based Motion Estimation for Image Reconstruction Using Bi-direction Confidential
    (2014) Darun Kesrarat; Vorapoj Patanavijit; Vincent Mary School of Engineering
    In block-based motion estimation where the outcome of the motion vector (MV) is used to reconstruct the image, noise is one of the major problems that impact the quality of the performance in image reconstruction. There are several aspects to improve the quality of the reconstructed image but we focus on improvement of the accuracy in MV from existing block-based motion estimation algorithms when they applies our proposed model only without other any additional models. Because we would like to prove that our proposed model improves an accuracy of the MV that it leads to the better quality of the reconstructed image as a result. This paper presents robust block-based motion estimation where bidirection confidential model is applied over the existing blockbased motion estimation algorithm to improve the accuracy of the MV itself. In the experiment where we simulated several Additive White Gaussian Noise (AWGN) levels over several experiment sequences, we found that the proposed model improved the quality of the reconstructed image when it is applied over several existing block-based motion estimation algorithms. In our experiment, we evaluated the quality of reconstructed image by using Peak Signal to Noise Ratio (PSNR).
  • Item
    Myanmar Paper Currency Recognition Using GLCM and k-NN
    (2016-01) Hlaing, Khin Nyein Nyein; Gopalakrishnan, Anilkumar Kothalil
    Paper currency recognition depends on the currency note characteristics of a particular country. And the features extraction directly affects the recognition ability. Paper currency recognition is one of the important applications of pattern recognition. This paper aims to present a model for automatic classification of currency notes using k-Nearest Neighbor (k-NN) classifier that is the most important and simplest method in pattern recognition. The proposed model is based on textural feature such as Gray Level Co-occurrence Matrix (GLCM). The recognition system is composed of four parts. The skew correction of rotated image is first. The captured image is second preprocessing and the third part is extracting its features by using GLCM. The last one is recognition, in which the core is k-Nearest Neighbor classifier. Experimental results are presented on a dataset of 500 images consisting of 5 classes of currency notes which are 100 Kyat, 200 Kyat, 500 Kyat, 1000 Kyat, and 5000 Kyat notes. It is shown that a good performance can be achieved using k-NN classifier algorithm. The recognition system presented in this paper indicates that the proposed approach is one of the most effective strategies of identifying currency pattern to read its face value.
  • Item
    Experimental Study on Image Reconstruction from Spatial Correlation-based Optical Flow Motion Vector over Non Gaussian Noise Contamination using Reversed Confidential with Bilateral Filter
    (2016) Darun Kesrarat; Vorapoj Patanavijit; Vincent Mary School of Engineering
    In motion estimation, noise is a verity to degrade the performance in optical flow for determining motion vector. This paper examines the performance of noise tolerance model in spatial correlation-based optical flow for image reconstruction from motion vector where the source sequences are contaminated by non Gaussian noise. There are Poisson Noise, Salt & Pepper Noise, and Speckle Noise. In the experiment, several standard sequences in different styles are used and the applied combination model of reversed confidential with bilateral filter on spatial correlation-based optical flow is mainly focused to determined the best condition to apply this model with. The result in image reconstruction from motion vector is used in performance comparison with traditional noise tolerance models by using Peak Signal to Noise Ratio (PSNR) as a primary index for studying.
  • Item
    An Effective Test Case Selection for Software Testing Improvement
    (2015-11) Adtha Lawanna
    One problem of testing software is selecting the suitable test cases from the test suit regarding the size of the programs. If the size of selected test cases is big, then it can affect the whole performance of software development life cycle. Accordingly, it increases testing time and produce many bugs. Therefore, this paper proposes the improvement of software testing for selecting the appropriate and small number of test cases by considering the amounts of the functions modified, lines of code changed, and numbers of bugs produced after modifying programs. The reason of proposing the software testing improvement model is to prepare effective algorithm, while numbers of bugs are lower than the traditional methods. According to the experimental results, the size of the selected test cases by using the proposed model is less than Retest All, Random, and a Safe Test about 98.70%, 87.86%, and 84.67% respectively. Moreover, the ability of STI is higher than the comparative studies about 1-20 times regarding the number of bugs found after modifying a program.
  • Item
    An Effective Model for Case-Based Maintenance in Cased-Based Reasoning Systems
    (2015-11) Adtha Lawanna
    Case-based reasoning systems have been applied for machine learning, artificial intelligence, knowledge-based systems and other related fields in order to provide the right solution to the right problem regarding the four processes, which are the process to retrieve, reuse, revise, and retain cases. This paper focuses on the last process because it produces two main problems, which are the size of a case base increase and the ability of preserving the competency decreases. These critical issues are occurring when repeating the cycles of case-based reasoning. Consequently, the case-based maintenance methods are developed to handle the situations. Accordingly, this paper proposes an effective model for case-based maintenance in casebased reasoning systems to give the best results compared with random, utility, footprint, footprint and utility deletion including case addition algorithm. By running the seven comparative studies on ten datasets retrieved from the machine learning repository, especially to study the efficiency of each algorithm in terms of reducing the size of the case base by selecting the small number of case solutions and preserving the competency after the maintenance systems are applied. According to experimental results, the effectiveness of the proposed model for storing the number of case solution gives the lower size of a case base, when compared with the existing techniques about 34.34%-114.84%. Besides, the percentage of adapting solutions for the traditional methods are lower than the proposed model as about 1.12-6.64 times, including the percent solving problem is lower than the effective model approximately 4.73%-33.55%.