Ebook: New Trends in Multimedia and Network Information Systems

We want to present to our readers this new monograph on the current trends in Multimedia and Network Information Systems. It discusses a very broad scope of subject matters including multimedia systems in their widest sense, Web systems, and network technologies. The monograph also includes texts devoted to more traditional information systems that draw on the experience of the Multimedia and Network Systems. Each of the discussed research trends is considered from both theoretical and practical viewpoints. Imposing a clear-cut classification for such a diverse research area is not an easy task.

The challenge is even greater due to the fact that in this book we tried to focus on the most topical research work of scientists from all over the world. The studies are original and have not been published anywhere else. In our opinion the chapters represent the dominant advances in computer information systems. It is worth emphasizing, that in most cases the research work relies heavily on the achievements and techniques developed originally in the area of Artificial Intelligence.

As a result, we have divided the monograph content into four major parts:

1. Multimedia Information Technology.

2. Data Processing in Information Systems.

3. Information System Applications.

4. Web Systems and Network Technologies.

Each of the parts covers a couple of chapters on detailed subject fields that comprise the area of its title.

We do hope that we have managed to collect and systematize the scientific knowledge on such a diverse field. We will be very pleased if this book inspires the research community working on Multimedia and Network Information Systems. If so, it will have achieved the goal that motivated the authors, reviewers, and editors.

Aleksander Zgrzywa

Kazimierz Choroś

Andrzej Siemiński

The paper describes and compares content-based global and local feature image retrieval techniques, which can be applied to estimate similarity between colour images. The database of 150 colour images with manually specified model similarity of each pair was prepared. The comparison of the analyzed methods was based on examining model and calculated similarities, which should be alike.

A method for image identification based on the pupil size analysis during human-computer interaction will be presented. The data was gathered during the experiment on 3D interactive representation of fairly large picture collections which facilitates browsing through unstructured sets of icons or pictures. The data was gathered using ASL gaze-tracking device. A group of 18 users took part in the experiment. The obtained gaze tracking results have shown that recorded pupil size patterns may indicate the fact of the presented image identification.

In this chapter, the authors propose an algorithm for packet loss concealment (PLC) in transmission over IP-based networks with high packet loss rate. The algorithm is a sender-receiver-based extension of ANSI T1.521a Annex B PLC standard for G.711 voice codec. It consists in adding to a transmitted packet redundant parameters describing speech signal in another packet. Efficiency of the proposed algorithm was verified using subjective Absolute Category Rating (ACR) method and objective PESQ algorithm, and compared with original ANSI T1.521a Annex B standard. The intelligibility of speech was assessed using Semantically Unpredictable Sentences (SUS) tests. For high packet loss rates, all assessment methods proved superiority of the proposed algorithm over the original ANSI standard. The ACR tests showed that the proposed method can maintain speech quality above 3 in MOS scale even for packet loss rates of 20%-25%.

Visual retrieval systems as well as Internet search engines demand efficient techniques of indexing to facilitate fast access to the required video or to the required video sequences in the video databases. Digital video databases are more and more frequently implemented not only in the Internet network but also in local networks and even in local personal computers. Different approaches to digital video indexing are applied: textual approach, extraction and representation of technical or structural features, content-based analysis, and finally segmentation. The segmentation process leads to the partition of a given video into a set of meaningful and individually manageable segments, which then can serve as basic units for indexing. Video has temporal properties such as camera motion, object movements on the scene, sequential composition, and interframe relationships. An effective segmentation technique is able to detect not only abrupt changes but also gradual scene changes, such as fade and dissolve transitions. The nature of movies, mainly the style of video editing has an influence on the effectiveness of temporal segmentation methods. The effectiveness of four methods was analyzed for five different categories of movie: TV talk-show, documentary movie, animal video, action & adventure, and pop music video. The cuts have been recognized as well as cross dissolve effects. The tests have shown that the specific nature of videos has an important influence on the effectiveness of temporal segmentation methods.

News video is a very important video source. Caption in a news video can help us to understand the semantics of video content directly. A caption localization and extraction approach for news video will be proposed. This approach applies a new Multi-Frame Average (MFA) method to reduce the complexity of the background of the image. A time-based average pixel value search is employed and a Canny edge detection is performed to get the edge map. Then, a horizontal scan and a vertical scan on this edge map are used to obtain the top, bottom, left and right boundaries of the rectangles of candidate captions. Then, some rules are applied to confirm the caption. Experimental results show that the proposed approach can reduce the background complexity in most cases, and achieves a high precision and recall. Finally, we analyze the relationship between background variation of frame sequence and detection performance in detail.

Nowadays most people own several devices, like a notebook or a smartphone, for satisfying their mobility and flexibility needs. Without special arrangements a program, its execution state and saved files - the latter two are commonly called “session” - are confined to a physical host. Session Mobility enables the user to break this law of locality, and instead creates a relation between the session and himself. In this paper we present a novel system for supporting session mobility in various scenarios. To support an almost automatic session handover between different devices a mobile agent system has been used. The selection of target devices has also been automated based on the usage context, the device's capabilities and a rough estimation of the actual location. Based on the Session Mobility in mobiLe Environments platform (SMiLE) a use case has been realized where a video session is migrated from a notebook to a smartphone and vice versa.

Due to the huge amount of currently collected data, only computer methods are able to analyze it. Data Mining techniques could be used for this purpose, but most of currently used techniques discovering global patterns loose information about local changes. In this paper the new patterns are proposed: frequent events and groups of events in data stream. They have two advantages: information about local changes in distribution of patterns is obtained and the number of discovered patterns is smaller than in other methods. Described experiments prove that patterns give valuable knowledge, for example, in analysis of computer logs. Analysis of firewall logs reveals interest of user, its favourite web pages and used portals. By using described methods for analysis of HoneyPot logs, detailed knowledge about malicious code and time of its activity could be received. Additionally, information about infected machines IP addresses and authentication data is automatically discovered.

Web mining employs the techniques of data mining to extract information from the Web for a variety of purposes. The usual sources of data are the log files of WWW or proxy servers. The paper examines the possibility of using the local browser buffer for that purpose. The data that could be extracted from both types of logs are compared. It turns out, that despite its limitations the browser buffer is a rich source of unique data about user navigational habits and the properties of the fragment of the WWW that he/she visits. The cache contains the both the full body of a WWW object as well as the header control data sent by the server. Additionally the cache includes some basic information about the usage pattern of each object. Therefore it is possible to study the susceptibility to buffering the objects which is measured by the CF (cacheability factor) and to study the word diversity of Internet texts seen by the user. The CF factor provides an objective measure of the web site caching potential and thus makes it possible to infer about latency of the web site. The word diversity study tests the compliance of the Internet texts with the well known Zipf and Heaps Laws' that are valid for all natural languages. That part study could be used for the optimization indexing engines or the recommendation of pages potentially interesting for the user.

Existing classification schemes are visualized as hierarchical trees. Science data visualization requires a new method in information space modelling in order to reveal relations between class nodes. This paper describes a novel visualization concept of classification scheme using subject content metrics. We have mapped the document collection of Association for Computing Machinery (ACM) digital library to a sphere surface. To overcome the incorrectness of linear measures in indexes distances we calculated similarity matrix of themes and multidimensional scaling coordinates. The results show that space distances between class nodes accurately correspond with the thematic proximities. Documents mapped into a sphere surface were located according to the classification nodes and distributed uniformly. Proposed method to visualize classification scheme is proper to reach nonlinearity in subject content visualization. This property allows us to place close by more classification nodes. Symmetry of a sphere favours a new subclasses and sublevels of classification trees uniform visualization. This method may be useful in the visual analysis of Computer Science and Engineering domain development being grown instantly.

Predictive Toxicology (PT) attempts to describe the relationships between the chemical structure of chemical compounds and biological and toxicological processes. The most important issue related to real-world PT problems is the huge number of the chemical descriptors. A secondary issue is the quality of the data since irrelevant, redundant, noisy, and unreliable data have a negative impact on the prediction results. The pre-processing step of Data Mining deals with complexity reduction as well as data quality improvement through feature selection, data cleaning, and noise reduction. In this paper, we present some of the issues that can be taken into account for preparing data before the actual knowledge discovery is performed.

This paper presents a new method for storage and access to spatiotemporal data. That is spatial objects that have some non-spatial attributes updated asynchronously. An example of such objects may be water meters. Proposed method is a hybrid of well documented dedicated solutions for spatial, temporal, spatial aggregate and temporal aggregate data processing. Thanks to that it was possible to achieve high performance for detailed and aggregate query processing without usage of approximation. Index name (i.e. STAH-tree) is English abbreviation and can be extended as Spatio-Temporal Aggregation Hybrid Tree. Part of this work aims in creation of cost model checked against experimental results of system performance. Some other experiments that verify system behavior were also performed.

Fuzzy models to assist with real estate appraisals are described and previous experiments on optimizing them with evolutionary algorithms implemented in MATLAB are summarized. An approach was made to use the KEEL Tool, developed in Java by a group of Spanish research centres, to investigate the models. Five fuzzy models comprising 3 or 4 input variables referring to the attributes of a property were learned and evaluated using six regression algorithms for fuzzy rule based systems implemented in the KEEL Tool. The experiments were conducted using a data set prepared on the basis of actual 134 sales transactions made in one of Polish cities and located in a residential section. The results were encouraging. The KEEL Tool has proved to be very useful and effective research tool, especially thanks to its 10-fold cross validation mechanism and relatively short time of data processing.

A detailed description of top-k spatial preference queries and the schema of their execution will be presented. Then, it discusses existing R-tree based top-k spatial preference queries' execution algorithms and the optimization methods they utilize. Moreover, the paper presents a new optimization technique and an algorithm capable of utilizing it together with other methods. Finally, an analysis of proposed method's efficiency is presented.

The digitisation of national heritage is supported in Poland by the government who want to establish standards and recommendations that could be used by all types of cultural institutions and organizations involved in digitisation process, including museums, libraries and archives. This paper describes an approach to establishing standards concerning technical aspects of digitisation, first of all technical and structural metadata and a set of rules, parameters and formats. The presented solutions are based on an unpublished report prepared by a working team.

This report presents a novel methodology in evaluating performance of the popular on-line multilingual search engine AltaVista. We study some crucial aspects of natural language that usually disrupt translation process and the extend to which it influences retrieval results. Having prepared the test set, we analyze phenomena of an English and French language pair in relation to the strategy of browsing the Web for documents with the features specified in the query. Using Natural Language Processing techniques, we test a Machine Translation system performance in order to improve the translation quality, which in turn has an impact on the results of information retrieval in a language other than the query language.

This article describes the approach to create the multi-agent system (MAS) for simulating transport corridors. There are few new ideas that were used during the developing the system. JASON language used in implementation of this system is a relatively new approach. A commitment based communication was described before in few papers but there is a little about its implementation. The social aspects of agents are emphasizes in this paper. Although the system is not fully implemented, the combination of Prometheus methodology and JASON language seems to be the good choice for developing the multi-agent systems.

E-learning systems are intended to offer their users didactic material tailored to their actual needs, preferences and abilities. This idea is consistent with the general principles of constructivism. Additionally it offers a student great flexibility in organizing his own curriculum and his learning scenario in the scope of particular course. However, it has also some side effects causing that students have many problems with making curriculum adapted to their needs or choosing subsequent course elements while complying the course. Many of e-learning courses designers, teachers and students emphasize the necessity of guidance in effective traversing a “knowledge network”. The chapter presented below explores some of the recent solutions in this subject. They were carefully selected to represent the widest possible range of research concerning the application of swarm intelligence to recommendation in e-learning systems.

The main purpose of intelligent tutoring systems is to guarantee an effective learning and offer the optimal learning path for each student. Therefore, determination of learning scenario is a very important task in a learning process. After a new student is registered in the system, he is classified to the appropriate group. Before he begins to learn an opening scenario is determined based on final scenarios of students who belong to the class of learners similar to the new one. The new student is offered the optimal learning path suitable for his preferences, learning styles and personal features. In this paper new knowledge structure, which involves version of lessons, is proposed. For the defined knowledge structure definitions of learning scenario, distance function and the procedure of the scenario determination are presented.

In this paper we will report on work in progress towards increasing the relational density of the German wordnet. It is also an experiment in corpus-based lexical acquisition. The source of the acquisition is a large corpus of German newspaper texts. The target is the German wordnet (GermaNet). We acquire a new type of lexical-semantic relation, i.e. the relation between the verbal head of a predicate and the nominal head of its argument. We investigate how the insertion of instances of this relation into the German wordnet GermaNet affects the neighbourhood of the nodes which are connected by an instance of the new relation. Special attention is given in this paper to the language-specific aspects of the acquisition process.

Cadastral systems belong to the most important public systems, since they provide necessary information for economic planning, spatial planning, and tax calculation, real estate denotation in perpetual books, public statistics, and real estate management as well as farm registration. They also provide source data to systems such as IACS and its main component LPIS. Quality of cadastral data has a major impact on the functioning many other systems. Legacy cadastral systems were developed using different data structures and various computer technologies. The quality of those legacy systems was sufficient for accomplishing everyday tasks but it was too low in the case of transferring data into other information systems. A practical approach to assure data quality in the process of transferring descriptive and geometric data from legacy systems into the Kataster OnLine - a modern integrated cadastral information system is presented in the paper.