Sule Yildirim Yayilgan

Semantic Tags for Lecture Videos

In an effort to develop effective multi-media learning objects (MLO), we propose a framework to extract and associate semantic tags to temporally segmented instructional videos. These tags serve for the purpose of efficient indexing and retrieval system. We create these semantic tags from potential keywords extracted from the lecture transcript. The keywords undergo a series of refinement process to select few but meaningful set of tags. We use word similarity measure using visual ness and word sense disambiguation to select the tags from candidate keywords. These tags are finally associated with video segments in which they appear based on timestamp. Each video segment represents a key idea or a topic. We also evaluated the objective keyword selection criteria to subjective test with some interesting results.

Pixelwise object class segmentation based on synthetic data using an optimized training strategy

In this paper we present an approach for low-level body part segmentation based on RGB-D data. The RGB-D sensor is thereby placed at the ceiling and observes a shared workspace for human-robot collaboration in the industrial domain. The pixelwise information about certain body parts of the human worker is used by a cognitive system for the optimization of interaction and collaboration processes. In this context, for rational decision making and planning, the pixelwise predictions must be reliable despite the high variability of the appearance of the human worker. In our approach we treat the problem as a pixelwise classification task, where we train a random decision forest classifier on the information contained in depth frames produced by a synthetic representation of the human body and the ceiling sensor, in a virtual environment. As shown in similar approaches, the samples used for training need to cover a broad spectrum of the geometrical characteristics of the human, and possible transformations of the body in the scene. In order to reduce the number of training samples and the complexity of the classifier training, we therefore apply an elaborated and coupled strategy for randomized training data sampling and feature extraction. This allows us to reduce the training set size and training time, by decreasing the dimensionality of the sampling parameter space. In order to keep the creation of synthetic training samples and real-world ground truth data simple, we use a highly reduced virtual representation of the human body, in combination with KINECT skeleton tracking data from a calibrated multi-sensor setup. The optimized training and simplified sample creation allows us to deploy standard hardware for the realization of the presented approach, while yielding a reliable segmentation in real-time, and high performance scores in the evaluation.

Adult Video Content Detection Using Machine Learning Techniques

Automatic adult video detection is a problem of interest to many organizations around the world. The aim is to restrict the easy access of underage youngsters to such potentially harmful material. Most of the existing techniques are mere extensions of image categorization approaches. In this paper we propose a video genre classification technique tuned specifically for adult content detection by considering cinematographic principles. Spatial and temporal simple features are used with machine learning algorithms to perform the classification into two classes: adult and non-offensive video material. Shot duration and camera motion, are the temporal domain features, and skin detection and color histogram are the spatial domain ones. Using two data sets of 7 and 15 hours of video material, our experiments comparing two different SVM classifiers achieved an accuracy of 94.44%.

Security and Privacy Considerations for IoT Application on Smart Grids: Survey and Research Challenges

The emergence and evolution of Internet of Things (IoT) offers great advantages to improve substantially the management over electricity consumption and distribution to the benefit of consumers, suppliers and grid operators. However, introducing IoT related devices and technologies in smart grids might lead to new security and privacy challenges. Though necessary technological innovations to ensure secure communication are being developed, more work is still required towards more secure standards for communication between devices and Smart Grids. This paper provides an overview about the security and privacy challenges of IoT applications in smart grids. Furthermore, we highlight and analyze some solutions and practices being used to cope with security and privacy requirements for IoT on deployment and management of smart grid. We address three types of challenge domains: customer domain, information and communication domain, and the grid domain.

SEMCON: Semantic and contextual objective metric

This paper proposes a new objective metric called the SEMCON to enrich existing concepts in domain ontologies for describing and organizing multimedia documents. The SEMCON model exploits the document contextually and semantically. The preprocessing module collects a document and partitions that into several passages. Then a morpho-syntatic analysis is performed on the partitioned passages and a list of nouns as part-of-speech (POS) is extracted. An observation matrix based on statistical features is then computed followed by computing the contextual score. The semantics is then incorporated by computing a semantic similarity score between two terms - term (noun) that is extracted from a document and term that already exists in the ontology as a concept Eventually, an overall objective score is computed by adding contextual score with semantic score. Subjective experiments are conducted to evaluate the performance of the SEMCON model. The model is compared with state-of-the-art tf*idf and χ2 (Chi square) using FI measure. The experimental results show that SEMCON achieved an improved accuracy of 10.64 % over the tf*idf and 13.04 % over the χ2.

An Improved Concept Vector Space Model for Ontology Based Classification

This paper proposes an improved concept vector space (ICVS) model which takes into account the importance of ontology concepts. Concept importance shows how important a concept is in an ontology. This is reflected by the number of relations a concept has to other concepts. Concept importance is computed automatically by converting the ontology into a graph initially and then employing one of the Markov based algorithms. Concept importance is then aggregated with concept relevance which is computed using the frequency of concept occurrences in the dataset. In order to demonstrate the applicability of our proposed model and to validate its efficacy, we conducted experiments on document classification using concept based vector space model. The dataset used in this paper consists of 348 documents from the funding domain. The results show that the proposed model yields higher classification accuracy comparing to traditional concept vector space (CVS) model, ultimately giving better document classification performance. We also used different classifiers in order to check for the classification accuracy. We tested CVS and ICVS on Naive Bayes and Decision Tree classifiers and the results show that the classification performance in terms of F1 measure is improved when ICVS is used on both classifiers.

Multimodal Biometric Authentication Using Fingerprint and Iris Recognition in Identity Management

The majority of deployed biometric systems today use information from a single biometric technology for verification or identification. Large-scale biometric systems have to address additional demands such as larger population coverage and demographic diversity, varied deployment environment, and more demanding performance requirements. Todays single modality biometric systems are finding it difficult to meet these demands, and a solution is to integrate additional sources of information to strengthen the decision process. A multibiometric system combines information from multiple biometric traits, algorithms, sensors, and other components to make a recognition decision. Besides improving the accuracy, the fusion of biometrics has several advantages such as increasing population coverage, deterring spoofing activities and reducing enrolment failure. The last 5 years have seen an exponential growth in research and commercialization activities in this area, and this trend is likely to continue. Therefore, here we propose a novel multimodal biometric authentication approach fusing iris and fingerprint traits at score-level. We principally explore the fusion of iris and fingerprint biometrics and their potential application as biometric identifiers. The individual comparison scores obtained from the iris and fingerprints are combined at score-level using a three score normalization techniques (Min-Max, Z-Score, Hyperbolic Tangent) and four score fusion approaches (Minimum Score, Maximum Score Simple Sum and User Weighting). The fused-score is utilized to classify an unknown user into the genuine or impostor.

Towards Understanding the MOOC Trend: Pedagogical Challenges and Business Opportunities

Undoubtedly, MOOCs have the potential to introduce a new wave of technological innovation in learning. In spite of the great interest among the educators and the general public MOOCs have generated, there are some challenges that MOOCs might face when it comes to examining and determining the best pedagogical approaches that MOOCs should be based on. Moreover, MOOCs are facing also challenges towards building a consistent business model. The main objective of this paper is to shed more light on the MOOCs phenomenon, by analyzing and discussing some benefits and drawbacks of MOOCs from the pedagogical and business perspectives. Therefore, in this paper we provide an in-depth analysis of MOOCs challenges and opportunities towards determining pedagogical innovations. We also analyze current trends of MOOCs expansion to create new educational markets by overpassing the bricks-and-mortar educational institutions. To do so, we conduct a SWOT analysis on MOOCs. Finally, we provide possible directions and insights for future research to better understand how MOOCs can be improved to lead to greater innovations in the higher education landscape to answer the needs of a knowledge-based economy.

Playful participation for learning in higher education — The introduction of participatory role play simulation in a course at Hedmark University College

A playful approach to supplement theoretical input in lectures, are role play simulations. The role play simulation tested out at Hedmark University College, diverts from the more generic form of role play simulations by its participatory approach. This paper presents the results from this complete testing of a methodology developed iteratively. The methodology consists of four stages, where each has been developed separately or in pairs.

An intelligent model for predicting the occurrence of skiing injuries

Artificial neural networks offer a unique way to model very complex and innovative systems that can be very effective in anticipating various accident severities. In this article, we propose a neural-network-based model, able to predict the number of severe injuries caused while skiing. The proposed system is intended for use by ski patrol and medical personnel to better prepare themselves in advance for treating ski-injured persons. The ski patrol and any other medical personnel will be able to know the statistics, type and severity of the injuries occurred, and most importantly, will be benefiting from having predictions for each day. Considering that, the number of injured people in a particular place each day was estimated, the results are very promising suggesting that such a system would prove beneficial in accurately predicting skiing injuries.