Multimedia and Network Information Systems

Consecutive stages of building knowledge for automatic lip speech identification are shown in this study. The main objective is to prepare audio-visual material for phonetic analysis and transcription. First, approximately 260 sentences of natural English were prepared taking into account the frequencies of occurrence of all English phonemes. Five native speakers from different countries read the selected sentences in front of three cameras. Video signals, synchronized with audio, were registered and then analyzed. Encountered problems related to video registration and results achieved are discussed.

This paper focuses on manners of filling the gaps in existing standards that are used in tridimensional Web technologies. We proposed the way of encoding huge 3D data sets in lossless PNG format and use of programmable rendering pipeline to decoding PNG file. It allows to reduce significantly the file with 3D data, time of transmission via Web and time needed to decode the file.

The image signature concept can be a successful approach to image comparison in content-based retrieval, but it is a very challenging task. Fundamental for this purpose is defining signature similarity. There exist a lot of similarity models which measure similarity of images or their objects in multimedia databases. In order to deal with semantic retrieval, we have introduced a three stage search engine. At each stage, we need different but equally effective similarity measures. Here, we have analysed asymmetric and symmetric approaches to signature comparison. In our experiment, we present an extensive comparison of some similarity measures dedicated to image retrieval.

Sports videos are the most popular videos searched in the Web. To retrieve efficiently and effectively sports videos we need to use very sophisticated techniques of content-based analysis. Many strategies of content-based indexing have been already proposed, tested, and applied to categorize video shots in sports news videos. One of the techniques is based on player scene analyses leading to the detection of playing fields. The characteristic of a playing field strongly depends on the sports category. Some sports videos are characterized by a very dynamic background, others by a static background, close-up view of players, in-field medium view, wide view, or out of field view of the audience, small or great objects of foreground, homogeneous type of playing field with one dominant color or very diversified field. The recognition of such sports video features as dominant color of playing field and type of shot view can significantly help to categorize sports news videos. The paper discusses some aspects of the processes of automatic dominant color extraction and playing field detection basing on the experiences achieved during the experiments performed in the Automatic Video Indexer AVI.

In multi-label classification tasks, very often labels are correlated and to not lose important information, methods should take into account existing dependencies. Such situation especially takes place in the case of multimedia datasets. In the paper, universal problem transformation methods providing for label correlations are considered. The comparison is done for proposed by authors Labels Chain technique [4] and well known methods which also take into account label correlations, such as Label Power-set, Classifier Chains and Ensembles of Classifier Chains. The performance of the methods is examined by experiments done on image, musical, audio and text datasets.

A statistical machine translation (SMT) capability would be very useful in augmented reality (AR) systems. For example, translating and displaying text in a smart phone camera image would be useful to a traveler needing to read signs and restaurant menus, or reading medical documents when a medical problem arises when visiting a foreign country. Such system would also be useful for foreign students to translate lectures in real time on their mobile devices. However, SMT quality has been neglected in AR systems research, which has focused on other aspects, such as image processing, optical character recognition (OCR), distributed architectures, and user interaction. In addition, general-purpose translation services, such as Google Translate, used in some AR systems are not well-tuned to produce high-quality translations in specific domains and are Internet connection dependent. This research devised SMT methods and evaluated their performance for potential use in AR systems. We give particular attention to domain-adapted SMT systems, in which an SMT capability is tuned to a particular domain of text to increase translation quality. We focus on translation between the Polish and English languages, which presents a number of challenges due to fundamental linguistic differences. However, the SMT systems used are readily extensible to other language pairs. SMT techniques are applied to two domains in translation experiments: European Medicines Agency (EMEA) medical leaflets and the Technology, Entertainment, Design (TED) lectures. In addition, field experiments are conducted on random samples of Polish text found in city signs, posters, restaurant menus, lectures on biology and computer science, and medical leaflets. Texts from these domains are translated by a number of SMT system variants, and the systems’ performance is evaluated by standard translation performance metrics and compared. The results appear very promising and encourage future applications of SMT to AR systems.

In the article, the authors presented the possibilities of using the Universal Image Quality Index (Q)—a popular measure for evaluation of digital image quality in order to identify users based on analysing their fingerprints with the use of a reference image. The applied quality measure is used both for analysing of fingerprints as well as in the process of synchronisation preceding the analysis.

This paper describes a simple novel compound random field model capable of realistic modelling the most advanced recent representation of visual properties of surface materials—the bidirectional texture function. The presented compound random field model combines a non-parametric control random field with local multispectral models for single regions and thus allows to avoid demanding iterative methods for both parameters estimation and the compound random field synthesis. The local texture regions (not necessarily continuous) are represented by an analytical bidirectional texture function model which consists of single scale factors modeled by the three-dimensional moving average random field model which can be analytically estimated as well as synthesized.

Ubiquitous Computing has enabled users to perform their computer activities anytime, anyplace, anywhere while performing other routine activities. Voice-based interaction often plays a significant role to make this possible. Presently, in voice-based interaction system communicates information to the user sequentially whereas users are capable of noticing, listening and comprehending multiple voices simultaneously. Therefore, providing information sequentially to the users may not be an ideal approach. There is a need to develop a design strategy in which information could be communicated to the users through multiple channels. In this paper, a design possibility has been investigated that how information could be communicated simultaneously in voice-based interaction so that users could fulfil their growing information needs and ultimately complete multiple tasks at hand efficiently.

Music Information Retrieval (MIR) area as well as development of speech and environmental information recognition techniques brought various tools intended for recognizing low-level features of acoustic signals based on a set of calculated parameters. In this study, the MIRtoolbox MATLAB tool, designed for music parameter extraction, is used to obtain a vector of parameters to check whether they are suitable for separation of selected types of vehicle-associated noise, i.e.: car, truck and motorcycle. Then, cross-correlation between pairs of parameters is calculated. Parameters for which absolute value of cross-correlation factor is below a selected threshold, are chosen for further analysis. Subsequently, pairs of parameters found in the previous step are analyzed as a graph of low-correlated parameters with the use of the Bron-Kerbosch algorithm. Graph is checked for existence of cliques of parameters linked in all-to-all manner related to their low correlation. The largest clique of low-correlated parameters is then tested for suitability for separation into three vehicle noise classes. Behrens-Fisher statistic is used for this purpose. Results are visualized in the form of 2D and 3D scatter plots.

Brain computer interfaces become more and more available, mainly due to cheaper and better technology. Some of the devices, like NeuroSky MindWave and Emotiv EPOC became an example of the affordable apparatus that may be exploited for game interaction. At the same time most of authors, exploring brain waves interactions paradigms in games, concentrate on professional EEG devices, equipped with even hundreds of sensors, assuring high quality encephalography measures, what obviously outperform technically simplified solutions. Thus the paper provides an analysis of MindWave and Emotiv applicability for selected game interaction tasks: moving the object, selecting one of a few possibilities and interaction with the help of dialogue system. A corresponding game environment experiments were performed and analysis of control paradigms was provided.

Like other many fields, telemedicine has benefited from pervasive and ubiquitous access to knowledge granted by the internet and mobile wireless. Remote monitoring and diseases management are considered as most fasted growing areas within this field. Thanks to mobile communication technologies, humans today are able to provide patient with w better quality life in critical and emergency situations. In such a scenario, the patient needs to reach as soon as possible the health care provider and/or medical institution. To do so, he needs to go by a road without obstacles. In different research reviews, rooting algorithms that have been presented are treating the shortest path, the nearest path. In this paper, we present a new pervasive system able to find the Safe Area without mobile obstacles and impediments based on an incremental algorithm called Safe Region algorithm.

Collaborative filtering approach is one of the most widely used in recommendation processes. The big problem of this approach is its complexity and scalability. This paper presents an effective method for movie recommendation based on collaborative filtering. We show that the computational complexity of our method is lower than one known from the literature, worked out by Lekakos and Caravelas (Multimedia Tools Appl 36(1–2):55–70 (2006), [10]).

The Traffic Flow Description is an option of the IP protocol that allows end-systems to describe generated traffic flows. Such description includes instantaneous values of transmitted data in a given time. The option enables intermediate systems to assure QoS based on dynamic resource allocation. In this paper an implementation of the Traffic Flow Description option for the Linux kernel is presented. The paper includes both the description of the option, proposed by the Author as the Internet Draft working document and detailed description of the prototype implementation of the proposed option in the Linux kernel. The implementation covers both improvements introduced to the current long term stable 4.1.20 version of the Linux kernel and two helper functions that enable the option to be set up easily. Tests show that the functionality of the prototype implementation complies with the specification of the option, given in the Internet Draft. Results of performance tests show that the prototype implementation is able to work as a part of the system of QoS assurance.

Measurement results of quality LTE connections are presented in this paper. Play operator was selected as the LTE service provider. The transmission bitrates and response times were measured for home network and also while driving with a speed up to 130 km/h. The downloading and uploading bitrates for home network were observed during two weeks and there was calculated how these bitrates have changed during the day. The percentage of samples having a specific bitrates was calculated. There is shown that the request response time depends on the transmission bitrate. LTE connection stability was observed while driving.

In this paper we propose an approach for discovering unused resources and utilizing them in on-line multiplayers games in P2P environments. The proposed SelfAid Network automatically deploys and manages services running on machines belonging to end-users and connects players using the discovered resources. SelfAid Network consumes only spare resources, following the trend of sharing economy.

Paper focuses on bringing the classic ACO (Ant Colony Optimization) for TSP (Travelling Salesman Problem) to Hadoop ecosystem. Classic ACO can be parallelized for efficiency. Especially today, with virtualization and cloud computing it is particularly easy to run ACO simulation on many nodes. However the distribution part adds an extra cost to an implementation of a simulation.

The paper presents a study on the efficiency measures of the Ant Colony Communities (ACC). The ACC is an approach to parallelize the Ant Colony Optimization algorithm (ACO). An ACC is made up of a Community Server that coordinates the work of a set Ant Colony clients. Each client implements a classical ACO algorithm. The individual colonies work in an asynchronous manner processing data sent by server and sending back the obtained results. There are many possible locations for the clients: the same computer as the server, computers of a local or wide area network. The paper presents a detailed description of concept the ACC and reports the study of the efficiency of the such Communities. The efficiency is measured by their power (the amount of data processed in a given period of time) and scalability—the efficiency of adding colony clients on the Community. The paper contains also the taxonomy of parallel implementations of the Ant Colony.

This work presents several scenarios used to identify security incidents based on the analysis of web server log files. The main goal of this work is to identify security events triggered by web robots which can be considered as dangerous or unwelcome. Analysis of all security incidents was based on archived web server log files which were collected from 03.03.2014 to 31.01.2015 and came from the real and fully functional environment, available at www.darmowe-obrazki.pl. All data were obtained automatically on a daily basis and analyzed using Advanced Web Statistics software.

Data structures and relations are becoming increasingly complex and difficult to assess and manage. Although automated rules and algorithms can be used for many data-mining tasks, there are still situations where human attention and insight is required to identify unexpected circumstances or unanticipated patterns. Presentation of large quantities of data has always been a challenging task. In this paper a method for representing large graph-based data sets is proposed to help users navigate through large clusters of data. The proposed method is based on a stereoscopic 3D visualization with special enhancements for a large multi node graph visualization. The stereoscopic projection allows for utilization of techniques that can draw users’ attention to particular regions of the graph. The method uses specially established node-node relations to calculate attention drawing factor values for each graph node.

In this article we report the initial results of experiments using HMEANT metric (semi-automatic evaluation metric used for scoring translation quality by matching semantic role fillers) on the Polish language. The metric is evaluated in the task of Machine Translation (MT) and in re-speaking quality assessment. GUI-based annotation interface was developed and with this tool (https://github.com/krzwolk/HMEANT-metric-for-Polish) evaluation was conducted practically by not IT-related personnel. Reliability, correlation with automatic metrics, language independence and time costs were analysed as well. Role labelling and alignment using GUI interface were done by two annotators with no related background (they were only instructed for about 10 min). The results of our experiments showed high inter-annotator agreement as far as role labelling was concerned and a good correlation of the HMEANT metric with human judgements based on re-speaking evaluation.

The multilingual nature of the world makes translation a crucial requirement today. Within this research we apply state of the art statistical machine translation techniques to the West-Slavic languages group. We do West-Slavic languages classification and choose Polish as a representative candidate for our research. The experiments are conducted on written and spoken texts, which characteristics are defined as well. The machine translation systems are trained within West-Slavic group as well as into English. Translation systems and data sets are analyzed, prepared and adapted for the needs of West-Slavic—* translation. To evaluate the effects of different preparations on translation results, we conducted experiments and used the BLEU, NIST and TER metrics. By defining proper translation parameters to morphologically rich languages we improve the translation quality and draw the conclusions.

As opposed to query modelling, relevance generating interactive query refinement (QR) is a technique aimed at exploiting syntax variations of gradually extended, being removed or replaced with some other keywords query, which depending on the factors like e.g. the information resource, the database structure, or the keyword alignment, facilitates significantly the searching process. Therefore our motivation is to explore the dynamism of the precision trend depended upon the factors analyzed. For a couple of language pairs which constitute multilingual settings, we develop a user-centred framework that imposes distributed search optimization. Our data set contains variety of query types submitted to some translingual distributed search systems that perform a number of syntax-based indexing. We construct a dynamism of precision elevation trend that indicates what factors intensify the relevance set of the system responses from a perspective of the user’s information need.

With the development of the Internet and satellite television, access to thousands of programs and messages in different languages became widespread. Unfortunately, even well educated people do not speak sufficiently in more than two or three foreign languages, while most know only one, and this significantly limits the access to this information. In this paper, we define requirements for an automated system for Accessing Multilingual Information and opinionS (AMIS) that will help in the understanding of multimedia content transmitted in different languages, with simultaneous comparison to counterparts in their native language user. The concept of understanding we use will provide access to any information, regardless of the language in which it is presented. We believe that the AMIS project can have a immense and positive impact on the integration and awareness of society in social and cultural terms.

This paper presents an improvement over the results of Wang et al. on query answering to IQ test questions using word embedding. The improvement comes from using the Glove method and renormalization of results using the best results of two leading methods. This latter approach combines knowledge coming from different corpuses.

Together with the increasing role of Internet in commercial activity growing intensity of marketing content is observed. Advertising clutter is interfering with web usability and is affecting processing of the editorial content by web users. Therefore, effective way to manage marketing content is needed. This problem can be solved by using a proper combination of multi-criteria decision-analysis methods. The presented research shows a unique approach to identify assessment model of tradeoffs between the editorial content and the intensity of marketing components. The fuzzy model is identified on the basis of the experiment with the use of eye tracker and a combination of PROMETHEE and COMET methods. As a result, we obtained the assessment model, which is a relation between a set of defined inputs and a set of permissible outputs with the property that each input is related to exactly one output (assessment). Therefore, this model can be used online to manage web systems with balance between editorial and commercial content.

The multilingual nature of the world makes translation a crucial requirement today. Parallel dictionaries constructed by humans are a widely-available resource, but they are limited and do not provide enough coverage for good quality translation purposes, due to out-of-vocabulary words and neologisms. This motivates the use of statistical translation systems, which are unfortunately dependent on the quantity and quality of training data. Such systems have a very limited availability especially for some languages and very narrow text domains. Is this research we present our improvements to current quasi-comparable corpora mining methodologies by re-implementing the comparison algorithms, introducing a tuning script and improving performance using GPU acceleration. The experiments are conducted on lectures text domain and bi-data is extracted from web crawl from the WWW. The modifications made a positive impact on the quality and quantity of mined data and on the translation quality as well and used the BLEU, NIST and TER metrics. By defining proper translation parameters to morphologically rich languages we improve the translation quality and draw the conclusions.

The named entity recognition (NER) problem has an important role in many natural language processing (NLP) applications and is one of the fundamental tasks for building NLP systems. Supervised learning methods can achieve high performance but they require a large amount of training data that is time-consuming and expensive to obtain. Active learning (AL) is well-suited to many problems in NLP, where unlabeled data may be abundant but labeled data is limited. The AL method aims to minimize annotation costs while maximizing the desired performance from the model. This study proposes a method to classify named entities from Tweet streams on Twitter by using an AL method with different query strategies. The samples were queried for labeling by human annotators based on query by committee and diversity-based querying. The experiments evaluated the proposed method on Tweet data and achieved promising results that proved better than the baseline.

Social networks represent nowadays an important communication channel for various groups of people over the world. They offer an audience for various activities with the aim to get the attention of the related target group, e.g. marketing or political campaigns. The aim of this paper is to understand the hid-den trends and various developments in such type of data and extract possible new and interesting knowledge for business purposes. For this purpose, we used data from two different social networks: Twitter and Tom’s Hardware. The completely analytical process was realised in line with CRISP-DM methodology; we selected the suitable methods of machine learning and exploratory data analysis to get the expected results. The created decision support application offers a group of methods to understand the data within the exploratory analysis, to generate a prediction model with the highest accuracy or to extract the rules supporting decision process during an on-line campaign. The best-achieved accuracy was higher than 95 % and extracted rules represent a good basis to ensure an expected popularity for selected topics in the future. Although we tested the system within a dataset closer oriented to the ICT sector, we will evaluate its applicability on a wider scale in our future work.

The paper presents a new online service, which provides an extensive support for financial analysis. The service is targeted at various groups of end-users, including small companies and their stakeholders, accountancy offices as well as financial advisors. The scope of the analysis involves five groups of financial ratios as well as percentage structure and dynamics analysis. The aim of the paper is to show the interdisciplinary and multidimensional nature of the service viewed from several different perspectives. The paper also aims at presenting certain programmatic solutions that we found useful and interesting while implementing this financial service.

One of the ways to describe the content of internet sources is known as topic modeling, which tries to uncover the hidden thematic structures in document collections. Topic modeling applied to social networks can be useful for analysis in case of crisis situations, elections, launching a new product on the market etc. It becomes popular research area in recent years and represents the methods to browse, search and summarize large amount of the textual data. The main aim of this paper is to describe a new way for topic modeling based on the usage of Formal Concept Analysis combined with reduction by Singular Value Decomposition of the input data matrix. In difference to other common used method for topic modeling our proposed method is able to generate topic hierarchy, which offer more detail analysis of topics within the collection. Our approach is experimentally tested on the selected dataset of Twitter network contributions.

In recent years, RDF event streams have been an increasingly widespread data source in a wide range of domains. Existing systems allow us to automatically record pieces of information concerning everyday fast-paced life. This paper proposes methods of representing and processing RDF streams consisting of RDF graphs with time-varying data annotated by an interval. We introduce a graph-based stream model and a solution for RDF streams processing based on the publish-subscribe interaction scheme. The initial evaluation on our implementation shows that it has great potential.

Gathering data from Internet of Things and management of IoT devices requires an efficient communication architecture. In this paper we analyse architectures of the scalable, sensor-oriented IoT network management system, as well as the pros and cons of introducing into it a message-oriented middleware server (message broker). We compare two architectures: with distributed buffers and with a centralized message broker. The analysis was conducted on the basis of Markov chains and discrete event simulation.

Although the history of distributed applications goes back to the 60s of the last century, the tremendous growth of opportunities faced by software developers has been in recent years. With the continuous increasing of access to the Internet, both in the traditional way using a desktop computer, and more often used mobile devices such as phones or tablets, the demand for providing more complex distributed systems increases. To meet their requirements it is necessary to introduce new solutions that will be able to handle a very large number of users from around the world. The paper presents the analysis of different ways of communication between distributed applications in .NET environment.

The main goal of an intelligent tutoring system is to provide learning materials suitable for students’ needs and preferences. Observations and analysis of students’ behaviour and interactions with the intelligent tutoring system are crucial to determine and, if necessary, modify the learning scenario. A properly designed user’s model influences the effectiveness of those methods and, in consequence, the whole learning process. This paper is devoted to propose a content of the student’s profile that includes behavioural measures.

Generalized One-Sided Concept Lattices (GOSCL) represent a tool for extraction of hidden hierarchical structure among the datasets with different types of attributes. The specific problem of this method is an interpretation of the results from large created hierarchies, what often leads to the selection of the most relevant concepts. Subsets quality measure and stability index are techniques used for the ranking of the concepts relevance. In this paper we describe an approach which combines these two ranking techniques. The proposed approach is illustrated by an example and the experiments with the effect of reduction on generated input data tables are also provided.