Ioannis Giannoukos

Dropout prediction in e-learning courses through the combination of machine learning techniques

In this paper, a dropout prediction method for e-learning courses, based on three popular machine learning techniques and detailed student data, is proposed. The machine learning techniques used are feed-forward neural networks, support vector machines and probabilistic ensemble simplified fuzzy ARTMAP. Since a single technique may fail to accurately classify some e-learning students, whereas another may succeed, three decision schemes, which combine in different ways the results of the three machine learning techniques, were also tested. The method was examined in terms of overall accuracy, sensitivity and precision and its results were found to be significantly better than those reported in relevant literature.

Features and classifiers for emotion recognition from speech: a survey from 2000 to 2011

Speaker emotion recognition is achieved through processing methods that include isolation of the speech signal and extraction of selected features for the final classification. In terms of acoustics, speech processing techniques offer extremely valuable paralinguistic information derived mainly from prosodic and spectral features. In some cases, the process is assisted by speech recognition systems, which contribute to the classification using linguistic information. Both frameworks deal with a very challenging problem, as emotional states do not have clear-cut boundaries and often differ from person to person. In this article, research papers that investigate emotion recognition from audio channels are surveyed and classified, based mostly on extracted and selected features and their classification methodology. Important topics from different classification techniques, such as databases available for experimentation, appropriate feature extraction and selection methods, classifiers and performance issues are discussed, with emphasis on research published in the last decade. This survey also provides a discussion on open trends, along with directions for future research on this topic.

Operator context scanning to support high segmentation rates for real time license plate recognition

Introducing high definition videos and images in object recognition has provided new possibilities in the field of intelligent image processing and pattern recognition. However, due to the large amount of information that needs to be processed, the computational costs are high, making the HD systems slow. To this end, a novel algorithm applied to sliding window analysis, namely Operator Context Scanning (OCS), is proposed and tested on the license plate detection module of a License Plate Recognition (LPR) system. In the LPR system, the OCS algorithm is applied on the Sliding Concentric Windows pixel operator and has been found to improve the LPR systems performance in terms of speed by rapidly scanning input images focusing only on regions of interest, while at the same time it does not reduce the system effectiveness. Additionally, a novel characteristic is presented, namely, the context of the image based on a sliding windows operator. This characteristic helps to quickly categorize the environmental conditions upon which the input image was taken. The algorithm is tested on a data set that includes images of various resolutions, acquired under a variety of environmental conditions.

Collaborative e-learning environments enhanced by wiki technologies

E-learning environments have met rapid technological advancements in the previous years. Nevertheless, current e-learning techniques do not adequately support student interaction and collaboration, resulting in decreased student progress and motivation. In this paper, a blended technique combining collaborative forums and wiki technologies is proposed. Through collaborative forums, students discuss course related topics assigned by the tutors to produce new educational material. This material is then stored in the wiki platform for further use. The proposed technique was applied on an e-learning course provided by the National Technical University of Athens and its effectiveness was evaluated using student activity data and questionnaire analysis. Results showed that the technique adequately supported teamwork, increasing student motivation and progress while simultaneously producing satisfactory level educational material.

Web-based decision-support system methodology for smart provision of adaptive digital energy services over cloud technologies

Energy information systems, which manage energy consumptions over internet, have been evolving over the past decade and can be considered as a part of a specialised sequential decision process, regarding the provision of personalised energy services to the community. The aim of this study is to develop and present an innovative decision-support system and cloud computing software methodology that brings together energy consultants, consumers, energy services procedures and modern web interoperable technologies. The authors propose a web-based knowledge system, using distributed cloud architecture and metering grids over ADSL broadband connections. By using some clustering algorithms and a web middleware, energy profiles over time are analysed and observed. The resulting clusters and centroids are projected and statistically analysed over time, producing a centroid-locus. Hypercube topology was used for efficient data management and software agent-based parallel analysis. The system operates efficiently on a multi-tier cloud-based middleware that generates in real-time using various service software components to the end consumers. The case study on real Greek energy measurements, for the first time in Greece, indicated a compact and efficient distributed procedure that could analyse and produce adaptive personalised information services.

Integrated vehicle recognition and inspection system to improve security in restricted access areas

Identification of vehicles for security reasons has lately attracted much scientific and commercial attention. In certain areas, such as government buildings, army camps or country borders, the vehicles are inspected before allowed to enter. As this inspection needs to be thorough, it is a rather time-consuming process. To address this issue, this paper proposes a combination of distinct computer vision inspection systems that have been developed in the Multimedia Technology Laboratory of the National Technical University of Athens, to automatically identify a vehicle entering a restricted area. The proposed system includes automatic license plate recognition, vehicle manufacturer/model detection and under-vehicle inspection. Combining the aforementioned methods can help to minimize the limitations of each individual technique and increase security in vehicle inspection. The characteristics, extracted during inspection, can be compared to the actual characteristics of each vehicle, which are acquired off-line and then stored in a vehicle record database.

Classification Models for Alzheimer’s Disease Detection

This paper presents a classification fusion for Alzheimer’s Disease (AD) and Mild Cognitive Impairment (MCI) classification based on dataset acquired basically from an automated structural MRI image processing pipeline. The dataset includes eighty-one regional cortical volume and cortical thickness features produced by the automated pipeline, along with two demographic measurements and three manual volume measurements of the hippocampus. This high-dimensional pattern classification problem is tested in a large database that contains clinical tests from six medical centers in Europe. The assessment of the results has shown that with a careful selection of combined classifiers, subject classification in three classes (Normal Controls, patients with MCI or with AD) is fairly accurate and can be used as an assistive tool to clinical examinations.

Integrating RFID on event-based hemispheric imaging for internet of things assistive applications

Automatic surveillance of a scene in a broad sense comprises one of the core modules of pervasive applications. Typically, multiple cameras are installed in an area to identify events through image processing techniques, which however present limitations in terms of object occlusion, noise, lighting conditions, image resolution and computational cost. To overcome such limitations and increase recognition accuracy, the video sensor output can be complemented by Radio Frequency Identification technology, which is ideal for the unique identification of objects. In this paper we examine the feasibility of integrating RFID with hemispheric imaging video cameras. After a brief description and discussion of related research regarding RFID location, video surveillance and their integration, we examine the factors that would render such a system feasible in terms of hardware, software and their environments. The advantages and limitations of each technology and their integration are also presented to conclude that their combination could lead to a robust detection of objects and their interactions within an environment. Finally, this work ends with the presentation of some possible applications of such integration.

An Adaptive Mechanism for Author-Reviewer Matching in Online Peer Assessment

Peer assessment techniques are an effective means to take advantage of the knowledge that exists in web-based peer environments. Through these techniques, participants act both as authors and reviewers over each other’s work. However, as web-based cooperating environments continuously grow in popularity, there is a need to develop intelligent mechanisms that will retrieve the optimal group of reviewers to comment on the work of each author, with a view to increasing the usefulness that these comments will have on the author’s final result. This paper introduces a novel technique that incorporates feed forward neural networks to determine the optimal reviewers for a specific author during a peer assessment procedure. The proposed method seeks to match author to reviewer profiles based on feedback regarding the usefulness of reviewer comments as it was perceived by the author. The proposed mechanism is expected to improve the peer assessment procedure, by making it adaptive to individual user characteristics, increasing the quality of the projects of a group overall and speeding up the peer assessment procedure. The method was tested on educational data derived from an e-learning course and the preliminary results that it yielded are promising.

An Intelligent Mechanism for Adaptive Peer User Modeling in Web-Based Environments

Peer assessment techniques are an effective means to take advantage of the knowledge that exists in Web-based peer environments. Through these techniques, participants act both as authors and reviewers over each other¿s work. However, as Web-based cooperating environments continuously grow in popularity, there is a need to develop intelligent mechanisms that will retrieve the optimal group of reviewers to comment on the work of each author, with a view to increasing the usefulness that these comments will have on the author¿s final result. This paper introduces a novel technique that incorporates feed forward neural networks to determine the optimal reviewers for a specific author during a peer assessment procedure. The proposed method seeks to match author to reviewer profiles based on feedback regarding the usefulness of reviewer comments as it was perceived by the author. The method was tested on educational e-learning data and the preliminary results that it yielded are promising.