ICITR - 2019
Permanent URI for this collectionhttp://192.248.9.226/handle/123/14731
Browse
Recent Submissions
- item: Conference-Full-textTeachers perception on learner performance when introducing ict to junior grades in schools: a case of Colombo district in Sri Lanka(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2019-12) Shashipriya, KWI; Sandanayake, TC; Sudantha, BHIt has been proved that the role of Information and Communication Technology (ICT), plays a vital role in the education sector in the process of enhancing and empowering the technology into the teaching, learning and assessment activities. One of the key objectives of teaching ICT in the local curriculum is to provide the prospects and trends of integrating technology enabled learning into the general educational activities. Enabling ICT in teaching, learning and assessment uplifts value entire education system, by enhancing the quality of learners. This study was aiming on the analyzing on learner analysis on students' performance when introducing ICT to junior grades in schools of Sri Lanka. The study has conducted aiming the teachers in Colombo district covering the government, semi-government and private schools. One of the great difficulties in this intervention is to get the trained teachers across the country to teach ICT for junior kids. Therefore this research study has aimed to identify the prevailing problems occurred during the process and to identify the possible avenues to overcome those problems. The research outcomes will be benefited to uplift the quality of learning amongst the students, parents and teachers leading to an effective use of ICT in junior grade students in the country.
- item: Conference-Full-textObject detection with deep learning for underwater environment(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2019-12) Wang, CC; Samani, H; Yang, CY; Sudantha, BHIn this research we have investigated the usage of deep learning algorithms for object detection in underwater environment and specifically we have employed YOLOv3 algorithm in our study. Details of the algorithm and experimental results are presented. We used available underwater database for training and investigated the method by experimenting to detect and identify the type of the fish in an aquarium in the lab. The results are also explained in this paper.
- item: Conference-Full-textiot driving assistant system for elderly(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2019-12) Huang, KM; Hsieh, TL; Yi, CA; Samani, H; Yang, CY; Sudantha, BHNowadays in this modern globalizing world, elders are given a more prominent place in the society. Thus a lot of new technological innovations are being invented in order to facilitate their day to day activities. The elder driving assistive technology is one such big step which has introduced to cater the demands of elderly people. This technology is beneficial in increasing safety of the driver and passengers, and also has affect in reducing the public cost of society safety. This study proposes a driver assistive tool for elder drivers which respond immediately in case of potential accidents by giving appropriate warning or arrestments. This tool monitors the movement of the vehicle dynamically with the behavior of the driver, detects the irregularity between the driver and the vehicle. Sensory devices which have been installed in the vehicle include the imaging camera, inertial measurement unit, Lidar scanner, steering wheel angle sensor, depression sensors on accelerator pedal and brake pedal to form a sensory network for the purpose of collecting signals for irregularity identification. In order to detect the surrounding suspicious objects, an array of ultrasonic sensors was installed on the vehicle and to evaluate the irregularity level of the danger, a sensor harness was integrated. When the risk level reaches a significant level, a danger classification information will be delivered to the processing center and a corresponding ensemble of sensory feedback will be activated to remind to the driver. Driver will be notified via the signal interface automatically, which has been implemented to resist the operation or even stop the vehicle immediately when a dangerous situation related to inaccurate behavior is detected to prevent potential disasters.
- item: Conference-Full-textDetecting and capturing the intensity of a brain tumor using mri images(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2019-12) Gayanga, TDL; Pathirana, GPSN; Sandanayake, TC; Sudantha, BHMutated cells in the brain are known as tumors. There are two types of tumors present in the brain, and they are malignant and benign tumors. Benign tumors are non-cancerous tumors which reside inside a person's brain having a primitive shape and size. Malignant tumors are cancerous brain tumors which spread by transforming the cells into the cell type next to the malignant tumor and have no clearly defined edge or shape(cloudy). Treatment planning and detection is the most efficient way of treating a patient and improving the condition. Magnetic Resonance Imaging (MRI) is the established method to detect the tumors. After the detection classification of the tumor manually takes reasonable time and sometimes it is impossible to detect the type with the naked eye. This results in having to conduct a biopsy which is a risk as well as sometimes impossible. One of the main factors to be considered in segmenting the tumor is the edge intensity of the tumor. For detecting the intensity of the tumor edge a novel completely automatic and reliable detection based on CNN is proposed. Pre-processing and running the improved images through CNN to get the best possible features to be extracted from the MRI images. Canny edge detection and Wavelet transform are applied to detect the edge and finally, Hough transform is used to detect the intensity of the edge. Edge detection method that proposed in this paper will have a wide verity of surgical applications.
- item: Conference-Full-textSpeech document summarization using neural network(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2019-12) Balasundaram, K; Amalraj, CRJ; Sudantha, BHText summarization is important to reduce content overloaded. Our aim of the automatic text summarization is extract the main content from the speech in a meeting or conference for the document purpose. It is difficult to remember all the things by human. At the same time when we write about the meeting/conference there is possibility to miss the content because of some inconvenience like noises, become lazy by the long speech, getting tired and other distraction . We have proposed an automatic system which can be help to get the summarized document. At First, we translate the recorded speech in to text document using google speech recognition API. Then summarization is done using that text document.
- item: Conference-Full-textSpeaker change detection for conversational speech using synthesized voice embedding(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2019-12) Krishnathasan, M; Amalraj, CRJ; Sudantha, BHSpeaker change detection is a prominent area of research in voice tasks for many years but used in very limited areas. Speech segmentation, feature extraction, and classification techniques are used as pipeline modules to detect speaker change in speech signals like conversations. This paper provides an approach to detect speaker changes in a conversational speech. Along with that, this paper demonstrates experiments on various sub modules and hyper parameters in the proposed pipeline.
- item: Conference-Full-textFeature based speaker embedding on conversational speeches(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2019-12) Balasubramaniyam, H; Amalraj, CRJ; Sudantha, BHIdentifying the speaker of a specific speech by examining the speech features of the voice is called speaker identification. The task of speaker identification consists of three main phases which are feature extraction, feature embedding and voice classification. Speaker embedding is the process of modeling the voice of a person where the model of the utterance can uniquely represent the speaker of that voice. Speaker embedding is a commonly used method in Automatic Speaker Recognition systems to identify the voice of the speaker. Currently, Deep Neural Networks based approaches are used in these systems for speech feature extraction and speech embedding. The performances of different approaches heavily depend on the noise factor and suitability of selected features of the audio data. MFCC, LPC, Dimensional filter banks are some of the frequently used speech features in speaker recognition. This speaker recognition research focuses on the usage of speech features for speaker embedding that are fitting for the speaker identification in conversational environment using a Convolutional Neural Network based approach.
- item: Conference-Full-textInvestigating the learning progress of cnns in script identification using gradient values(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2019-12) Tomioka, E; Morita, K; Shirai, NC; Wakabayashi, T; Ohyama, W; Sudantha, BHDemands for an automatic translation based on Camera-based Multilingual Optical Character Recognition (CM-OCR) are increasing. In addition, CM-OCR methods usually employ a script identification step before character recognition. Recent approaches for script identification depend on a Convolutional Neural Networks (CNN) thanks to its promising performance in the image recognition task. However, researchers mentioned the importance to understand the decision criteria in CNNs as a warning to employ them for actual tasks as black-box classifiers. Thus, the purpose of this research is to investigate the hyperparameter dependence of CNNs and to visualize the region focused by CNNs in the task of script identification. In this research, we applied Grad-CAM to the script identification task of image classification and used the SIW-13 dataset. We investigated the learning progress of CNNs by defining the value used in Grad-CAM as the "reaction" and visualized the region focused by CNNs in script identification. As a result, the learning process was stabilized in the case that the number of hyperparameters was sufficient for the given training samples even though the hyperparameters which should be tuned were increased. This result demonstrated that the capacity to stably learn training samples depends on the number of hyperparameters. In the insufficient capacity case, the learning process was destabilized and it caused scripts with relatively low accuracy. Analyzing one of the low accuracy scripts of the model using Grad-CAM, we found that some failures progress greatly changes by the difference in hyperparameters of CNNs. Scatter plots of the reaction and the probability clarified the capacity of CNNs in each script.
- item: Conference-Full-textAutomatic classification of neonatal sleep-wake states based on facial video analysis(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2019-12) Mukai, Y; Morita, K; Shirai, NC; Wakabayashi, T; Shinkoda, H; Matsumoto, A; Noguchi, Y; Shiramizu, M; Sudantha, BHPremature babies are admitted to the NICU (Neonatal Intensive Care Unit) for several weeks and generally placed under high medical supervision. To provide a better environment to them, some researchers investigate the affection of light and noise in the NICU on the formation of the sleep-wake cycle of the newborn called Circadian rhythm. These researches require the optimal evaluation method of the sleep-wake state. The visual assessment by nurses do not guarantee enough inter-tester reliability, and the measurement puts an additional burden on them. The conventional sleep-wake states discrimination method requires attachment devices on the subject's body. This paper proposes the automatic classification method of the sleep-wake states of neonates by using only facial information. In this research, we extract gradient features and spatio-temporal HOGV features from 3,600 face image frames (1 minute). According to Blazelton's method, this study classifies the sleep-wake states into six classes by using machine learning techniques. Support Vector Machine and Random Forest were used in the experiment. The spatio-temporal HOGV feature is an extension of the HOG feature to the time domain. The experiments using two kinds of feature quantities and classifiers showed that the highest accuracy rate (54.4%) was obtained by the gradient feature and Random Forest. This result suggested the possibility of improving accuracy by combining facial information with body movement and other conventional features.
- item: Conference-Full-textBlockchain based decentralized knowledge sharing system - jigsaw(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2019-12) Azeem, A; Jajeththanan, S; Sharmilan, S; Sudantha, BHKnowledge is formalized information that is cited or used in logical inference. The growth of knowledge sharing in the field of education helps people understand and grow their potential. Individuals are willing to share knowledge when they are certain to benefit from reciprocity or build a reputation. Individuals spend enormous amounts of time, resources and scarifications to gain knowledge. Individuals are currently sharing their expertise through forums, vlogs, videos, and trainings. Nevertheless, there is a variety of untrusted and unvalidated knowledge available for a subject from the recipient's point of view. To grasp what they need, it requires multiple resources. There are no direct benefits for people who make, comment or vote for these resources. In this research, the researcher implemented a distributed knowledge sharing system centered on Blockchain to allow multiple individuals to contribute their knowledge by building a resource that is moderated, verified, and community structured. Each creator, commenter and voter of knowledge will receive rewards, and the knowledge invested will gain it forever. The researchers used cryptography and knowledge economy in addition to stellar blockchain to make this process secure and more trustable. In this paper, the proposed system is realistic, efficient and has an ever-earning model.
- item: Conference-Full-textA review of query optimization techniques for complex event processing(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2019-12) Perera, KPD; Ahangama, S; Sudantha, BHComplex Event Processing (CEP) Systems is a technology that is used in many modern fields of application such as finance and business analysis. It tracks and analyses data from large information streams pertaining to a string of related or non-related events in order to identify patterns and relations that could be used to derive useful connections among seemingly unrelated factors within its applications. CEP systems make use of pattern queries to match identified events within an event stream. However, due to the generalized nature of CEP query languages and the lack of general structure and semantics, it is difficult to write queries that function optimally to deliver the expected results within the required time frames. This issue is particularly of importance as CEP systems often deal with time sensitive data and hence require rapid processing in order to output useful information and hence, defines the importance and requirement for query optimization techniques that may be applied to CEP systems. This paper focuses on research publications related to the four main pattern query optimization techniques, namely, Multi-Query Optimization, Join Query Optimization, Nested Query Processing techniques and Query Rewriting as well as their applications within modern CEP systems. This study further aims to identify possible limitations within the four techniques mentioned previously and advise on possible measures that may be taken to further improve these techniques in order to offer greater efficiency and stability to pattern query processing within CEP systems.
- item: Conference-Full-textA scenario-based er diagram and query generation engine(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2019-12) Hettiarachchi, S; Sugandhika, C; Kathriarachchi, A; Ahangama, A; Weerasuriya, GT; Sudantha, BHDesigning and developing a database is a crucial task in the System Development Life Cycle (SDLC). To design a database, it is essential to have proper knowledge about drawing Entity-Relationship (ER) Diagrams. Drawing ER diagrams is challenging for novices and people without a technical background. Furthermore, to retrieve data from a database requires expert domain knowledge about a database querying language like Structured Query Language (SQL). To address these issues, a system is proposed to identify and extract the necessary information from a given scenario to automatically generate the ER diagram. Based on that ER diagram, the system creates the database and is capable of generating SQL queries for any given type of natural language queries, in order to simplify accessing the data stored in the database.
- item: Conference-Full-textBlockchain & machine learning based secure personal medical record storage and sharing platform - datablock(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2019-12) Sharmilan, S; Farook, C; Sudantha, BHData is the most important part of machine learning. In the bioinformatics field, the sensitivity of the data is high and due to that, the accessibility of the data for a secondary purpose (e.g.: research) consists of many legal and ethical issues. Due to that in many bioinformatics research collecting the data consume more time than the development phase. There are some researches done to solve the legal and ethical issues by anonymizing the data using encryption, de-identification and perturbation of potentially identifiable attributes. But for some extend those solutions restricted the data breach but on the other hand, anonymized data not performed well during the analysis and mining tasks and some researches done to generate fake data like the real data sets. But those researches not full fill the requirements because of the generated data more restricted to the knowledge of the training data. The evolution of Blockchain provided a secure and trusted way to transfer valuable assets between two unknown parties. This research used Blockchain technology to store and share personal medical data to data scientists. And that will help them to build more accurate and efficient models. It also proposed a machine learning model to predict the authenticity and validity of the personal data based on domain knowledge and the validator's trust percentage.
- item: Conference-Full-textInteractive solution to improve flood awareness among public – flood run(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2019-12) Karunanayake, T; Dayarathne, P; Doratiyawa, C; Wickramanayake, A; Rankothge, WH; Gamage, N; Sudantha, BHFlood is the most common natural disaster in Sri Lanka that causes a huge destruction annually [1]. Lack of awareness on flood among the public is one of the main reasons behind the huge destruction of lives and property.We have proposed a platform for flood awareness, named "Flood Run" which is an interactive 3D mobile game, with following modules: puzzle games, action games and memorizing games, adventure games and quizzes. These four modules consist of activities to develop the essential skills to improve flood awareness. This research used game-based learning and interactive game designing techniques. This research paper presents the performance evaluation of the four types of modules. The results show that, with the help of the provided solution, the expected skills of the people are improved, and through that flood awareness among public is improved.
- item: Conference-Full-textDetecting automatically generated tweets using lexical analysis and profile credibility(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2019-12) Wickramarathna, NC; Ganegoda, GU; Sudantha, BHEvery industry relies heavily on accurate news and information distribution. In the recent decade social media has become one of the main methods of sharing news and other social impacting information online. But there's an uprising threat to all social media platforms, especially for twitter, known as bots. Not all bots are malicious but these automated accounts are largely responsible for platform manipulation, which is the process of misleading, disrupting the experience of other users by engaging in deceptive, aggressive activities. There are many politically motivated groups in Facebook and Twitter who use various levels of manipulation to influence voters and thereby undermining the democratic process. Platform manipulation is not only carried out by malicious automation, but also with spam and inauthentic accounts (fake accounts). This paper presents novel methodology to detect these bots (automated accounts) using existing research as foundation and builds new research solution to the problem. This methodology can be applied to the news domain to find bots involved in spreading false information. This methodology classifies a given tweet into either fake news or not and use the result as a feature and in addition to that user credibility can also be taken into account.
- item: Conference-Full-textA mobile application for crowdsourced road condition monitoring(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2019-12) Thilakarathna, TKK; Perera, HE; Jayaweera, HHE; Sudantha, BHAn Android App has been developed to measure geo-tagged vehicle-induced vibrations with the use of the built-in accelerometer of a mobile phone. The architecture of the app is designed in such a way that it can be published as a crowdsourced vehicle-induced vibration measurement data collection app. A simple calibration method has been developed to calibrate each individual axis of the accelerometer to get rid of imperfections within the three axes. The system is evaluated for extracting data of vibrations for a known system (simple pendulum) and the same system was tested for vehicle induced-vibration capturing. The results produced from the system during the test run correctly classified different road conditions. Once the crowdsourced data is available, a more statistically significant prediction could be performed.
- item: Conference-Full-textChange detection and tracking using synthetic aperture radar videos(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2019-12) Maithree, H; Dinushka, D; Wijayasiri, A; Sudantha, BHIn this paper, a change detection technique which can be utilized in identifying important changes in SAR videos has been proposed. Even though most researches have done on change detection in multi temporal SAR images, implementing a methodology to identify changes in SAR videos in a near real time manner has unique challenges such as the inherent speckle noise which will increase the false positive rate of the detection, rotation of the reconstructed SAR video frames due to the movement of the airborne vehicle, dynamic background and the overlapping of the area in consecutive video frames, non-uniform backscattering of SAR pulses and shadowy modelling of objects in video frames which doesn't provide much information about the appearance model of the objects. We propose an algorithm based on combination of optical flow calculation using Lucas Kanade method(LK method) and blob detection to detect the changes, and we feed the detected changes along with optical flow calculations to a centroid tracking algorithm to track the detected changes throughout the video.
- item: Conference-Full-text4onse as a complementary to conventional weather observation network(2019-12) Sudantha, BH; Warusavitharana, EJ; Ratnayake, GR; Mahanama, PKS; Warusavitharana, RJ; Tasheema, RP; Cannata, M; Strigaro, D; Sudantha, BHOver the centuries, sensing the weather has been important to mankind for decision making. Weather observations are important to better understand the climate variability and its consequences which ranges over different temporal and spatial scales such as localized rainfall and thunderstorms to large scale storms and droughts. Analysis of climate induced phenomenon is data intensive and the data collected from very sparse network of professional weather stations have become incapable to forecast magnitude of the climate induced events. In this research, we present a 4ONSE network as a complementary to professional, state-owned weather station network. This open network was built on 4 open pillars – open hardware, open software, open standards and open data. Credibility of the network was assessed by analyzing the reliability and accuracy of data, cost incurred in building the station and conformity of measured parameters with WMO standards. Even though the 4ONSE station is not a high tech, professional weather station, these mini stations can be used to filling the gaps left by professional weather stations. Thus, they can be used to improve the coverage of the existing weather network of the country and to obtain the observations at near real time for producing accurate weather and disaster forecasts.
- item: Conference-Full-textFeasibility evaluation of a solar powered automobile air-conditioning system(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2019-12) Ranaweera, WMMS; Nadhira, KF; Rathnayake, RPL; Alahakoon, PMK; Kumara, WGCW; Sudantha, BHThe rising demand for energy growth worldwide is a major crisis in the present. Development of the renewable energy creates new efficient solutions to existing applications. Currently, solar power is the best renewable energy source. Air-conditioning provides comfort living in sub-tropical countries like Sri Lanka. Solar-powered air-conditioning will reduce the operating costs. Therefore, this paper focuses on design and construction of a solar-powered air-conditioner for automobiles. Implementation of the solar powered air-conditioner in an automobile, will increase the fuel efficiency while reducing the carbon emissions. Further, by isolating the compressor from the engine and powering through the solar energy, engine load decreases. Experimental calculations are presented considering a 35 seater bus.