Skip to main content
main-content

Über dieses Buch

This book constitutes the proceedings of the 10th International Conference on Active Media Technology, AMT 2014, held in Warsaw, Poland, in August 2014, held as part of the 2014 Web Intelligence Congress, WIC 2014. The 47 full papers presented were carefully reviewed and selected from numerous submissions. The topics of these papers encompass active computer systems; interactive systems and applications of AMT-based systems; active media machine learning and data mining techniques; AMT for the semantic web; social networks and cognitive foundations.

Inhaltsverzeichnis

Frontmatter

Regular Contributions

Active Media and Ubiquitous Learning Foundations

A Semi-supervised Learning Algorithm for Web Information Extraction with Tolerance Rough Sets

In this paper, we propose a semi-supervised learning algorithm (TPL) to extract categorical noun phrase instances from unstructured web pages based on the tolerance rough sets model (TRSM). TRSM has been successfully employed for document representation, retrieval and classification tasks. However, instead of the vector-space model, our model uses noun phrases which are described in terms of sets of co-occurring contextual patterns. The categorical information that we employ is derived from the Never Ending Language Learner System (NELL) [3]. The performance of the TPL algorithm is compared with the Coupled Bayesian Sets (CBS) algorithm. Experimental results show that TPL is able to achieve comparable performance with CBS in terms of precision.

Cenker Sengoz, Sheela Ramanna

Identifying Influential Nodes in Complex Networks: A Multiple Attributes Fusion Method

How to identify influential nodes is still an open hot issue in complex networks. Lots of methods (e.g., degree centrality, betweenness centrality or K-shell) are based on the topology of a network. These methods work well in scale-free networks. In order to design a universal method suitable for networks with different topologies, this paper proposes a Multiple Attribute Fusion (MAF) method through combining topological attributes and diffused attributes of a node together. Two fusion strategies have been proposed in this paper. One is based on the attribute union (FU), and the other is based on the attribute ranking (FR). Simulation results in the Susceptible-Infected (SI) model show that our proposed method gains more information propagation efficiency in different types of networks.

Lu Zhong, Chao Gao, Zili Zhang, Ning Shi, Jiajin Huang

Local Regression Transfer Learning for Users’ Personality Prediction

Some research has been done to predict users’ personality based on their web behaviors. They usually use supervised learning methods to model on training dataset and predict on test dataset. However, when training dataset has different distributions from test dataset, which doesn’t meet independently identical distribution condition, traditional supervised learning models may perform not well on test dataset. Thus, we introduce a new regression transfer learning framework to deal with this problem, and propose two local regression instance-transfer methods. We use clustering and k-nearest-neighbor to reweight importance of each training instance to adapt to test dataset distribution, and then train a weighted risk regression model for prediction. We perform experiments on the condition that users dataset are from different genders and from different districts, and the results indicate that our methods can reduce mean square error about 30% to the most compared with non-transfer methods and be better than other transfer method in the whole.

Zengda Guan, Dong Nie, Bibo Hao, Shuotian Bai, Tingshao Zhu

Graph Clustering Using Mutual K-Nearest Neighbors

Most real world networks like social networks, protein-protein interaction networks, etc. can be represented as graphs which tend to include densely connected subgroups or modules. In this work, we develop a novel graph clustering algorithm called G-MKNN for clustering weighted graphs based upon a node affinity measure called ‘Mutual K-Nearest neighbors’ (MKNN). MKNN is calculated based upon edge weights in the graph and it helps to capture dense low variance clusters. This ensures that we not only capture clique like structures in the graph, but also other hybrid structures. Using synthetic and real world datasets, we demonstrate the effectiveness of our algorithm over other state of the art graph clustering algorithms.

Divya Sardana, Raj Bhatnagar

How Variability in Individual Patterns of Behavior Changes the Structural Properties of Networks

Dynamic processes in complex networks have received much attention. This attention reflects the fact that dynamic processes are the main source of changes in the structural properties of complex networks (e.g., clustering coefficient and average shortest-path length). In this paper, we develop an agent-based model to capture, compare, and explain the structural changes within a growing social network with respect to individuals’ social characteristics (e.g., their activities for expanding social relations beyond their social circles). According to our simulation results, the probability increases that the network’s average shortest-path length is between 3 and 4, if most of the dynamic processes are based on random link formations. That means, in Facebook, the existing average shortest path length of 4.7 can even shrink to smaller values. Another result is that, if the node increase is larger than the link increase when the network is formed, the probability increases that the average shortestpath length is between 4 and 8.

Somayeh Koohborfardhaghighi, Jörn Altmann

Big Data Management and Mining for Active Media

A Scalable Boosting Learner for Multi-class Classification Using Adaptive Sampling

Scalability has become an increasingly critical issue for successful data mining applications in the ”big data” era in which extremely huge data sets render traditional learning algorithms infeasible. Among various approaches to scalable learning, sampling techniques can be exploited to address the issue of scalability. This paper presents our study on applying a newly developed sampling-based boosting learning method for multi-class (non-binary) classification. Preliminary experimental results using bench-mark data sets from the UC-Irvine ML data repository confirm the efficiency and competitive prediction accuracy of the proposed adaptive boosting method for the multi-class classification task. We also show a formulation of using a single ensemble of non-binary base classifiers with adaptive sampling for multi-class problems.

Jianhua Chen

Scaling of Complex Calculations over Big Data-Sets

This article introduces a novel approach to scale complex calculations in extensive IT infrastructures and presents significant case studies in SONCA and DISESOR projects. Described system is enabling parallelism of calculations by providing dynamic data sharding without necessity of direct integration with storage repositories. Presented solution doesn’t require to complete a single phase of processing before starting the next one, hence it is suitable for supporting many dependent calculations and can be used to provide scalability and robustness of whole data processing pipelines. Introduced mechanism is designed to support case of still emerging data, thereby it is suitable for data streams e.g. transformation and analysis of data collected from multiple sensors. As will be shown in this article, this approach scales well and is very attractive because can be easily applied to data processing between heterogeneous systems.

Marek Grzegorowski

Music Data Processing and Mining in Large Databases for Active Media

The aim of this paper was to investigate the problem of music data processing and mining in large databases. Tests were performed on a large database that included approximately 30000 audio files divided into 11 classes corresponding to music genres with different cardinalities. Every audio file was described by a 173-element feature vector. To reduce the dimensionality of data the Principal Component Analysis (PCA) with variable value of factors was employed. The tests were conducted in the WEKA application with the use of

k

-Nearest Neighbors (

k

NN), Bayesian Network (Net) and Sequential Minimal Optimization (SMO) algorithms. All results were analyzed in terms of the recognition rate and computation time efficiency.

Piotr Hoffmann, Bożena Kostek

Mining False Information on Twitter for a Major Disaster Situation

Social networking services (SNS), such as Twitter, disseminate not only useful information, but also false information. Identifying this false information is crucial in order to keep the information on a SNS reliable. The aim of this paper is to develop a method of extracting false information from among a large collection of tweets. We do so by using a set of linguistic patterns formulated to correct false information. More specifically, the proposed method extracts text passages that match specified correction patterns, clusters the passages into topics of false information, and selects a passage that represents each topic of false information. In the experiment we conduct, we build an evaluation set manually, and demonstrate the effectiveness of the proposed method.

Keita Nabeshima, Junta Mizuno, Naoaki Okazaki, Kentaro Inui

Finding Cyclic Patterns on Sequential Data

The need for the study of dynamic and evolutionary settings made time a major dimension when it comes to data analytics. From business to health applications, being able to understand temporal patterns of customers or patients can determine the ability to adapt to future changes, optimizing processes and support other decisions. In this context, different approaches to Temporal Pattern Mining have been proposed in order to capture different types of patterns able to represent evolutionary behaviors, such as regular or emerging patterns. However, these solutions still lack on quality patterns with relevant information and on efficient mining methods. In this paper we propose a new efficient sequential mining algorithm, named PrefixSpan4Cycles, for mining cyclic sequential patterns. Our experiments show that our approach is able to efficiently mine these patterns when compared to other sequential pattern mining methods. Also for datasets with a significant number of regularities, our algorithm performs efficiently, even dealing with significant constraints regarding the nature of cyclic patterns.

António Barreto, Cláudia Antunes

Exploring Temporal Dependencies to Perform Automatic Prognosis

The use of data mining techniques in healthcare has been noticing an increased relevance over the last few years, being applied with a variety of objectives, with the most common one being the automatic diagnostic process. In this process, data mining techniques have achieved interesting and successful results. However, when it comes to prognosis the same quality of results is not being achieved. We argue that this happens thanks to the inability of the used techniques to capture the inherent temporal dependencies present on the data. Specifically, the temporal evolution of a patient is not being taken into account when performing prognosis. In this paper, we propose a different approach, independent of the domain, to address this issue. We present our preliminary results on two different datasets that show an improvement in the overall precision of the prognosis.

Daniel Cardoso, Cláudia Antunes

Active Media Engineering

Ryry: A Real-Time Score-Following Automatic Accompaniment Playback System Capable of Real Performances with Errors, Repeats and Jumps

In this work, we propose an automatic accompaniment playback system called Ryry, which follows human performance and plays a corresponding accompaniment automatically, in an attempt to realize human-computer concerts. Recognizing and anticipating the score position in real-time, known as score following, by a computer is difficult. The proposed system is based on a robust on-line algorithm for real-time audio-to-score alignment. The algorithm is devised using a delayed-decision and anticipation framework by modeling real-time music performance that includes uncertainties such as tempo fluctuation and mistakes. We developed an automatic accompaniment system that is capable of generating polyphonic music signals.

Shinji Sako, Ryuichi Yamamoto, Tadashi Kitamura

Towards Modular, Notification-Centric and Ambient Mobile Communication

User Study Supporting a New Interaction Model for Mobile Computing

In this paper, we synthesize the results of a qualitative user study supporting our proposed modular, notification-centric and ambient interaction model for mobile computing. Here, a context-dependent and extended implementation of notifications is introduced. For a range of personalized situations, individually tailored conceptual interfaces were compared against a baseline conform present-day mobile operating systems and the desired usage of suggested solutions as well as the perceived shortcomings of current offerings were investigated. We conclude and demonstrate the user preference for communication patterns introduced by our interaction model for mobile computing.

Jonas Elslander, Katsumi Tanaka

Body Posture Recognition as a Discovery Problem: A Semantic-Based Framework

The automatic detection of human activities requires large computational resources to increase recognition performances and sophisticated capturing devices to produce accurate results. Anyway, often innovative analysis methods applied to data extracted by off-the-shelf detection peripherals can return acceptable outcomes. In this paper a framework is proposed for automated posture recognition, exploiting depth data provided by a commercial tracking device. The detection problem is handled as a semantic-based resource discovery. A simple yet general data model and a corresponding ontology create the needed terminological substratum for an automatic posture annotation via standard Semantic Web languages. Hence, a logic-based matchmaking allows to compare retrieved annotations with standard posture descriptions stored as individuals in a proper Knowledge Base. Finally, non-standard inferences and a similarity-based ranking support the discovery of the best matching posture. This framework has been implemented in a prototypical tool and preliminary experimental tests have been carried out w.r.t. a reference dataset.

Michele Ruta, Floriano Scioscia, Maria di Summa, Saverio Ieva, Eugenio Di Sciascio, Marco Sacco

Literal Node Matching Based on Image Features toward Linked Data Integration

Linked Open Data (LOD) has a graph structure in which nodes are represented by Uniform Resource Identifiers (URIs), and thus LOD sets are connected and searched through different domains. In fact, however, 5% of the values are literal (string without URI) even in DBpedia, which is a

de facto

hub of LOD. Since the literal becomes a terminal node, and we need to rely on regular expression matching, we cannot trace the links in the LOD graphs during searches. Therefore, this paper proposes a method of identifying and aggregating literal nodes that have the same meaning in order to facilitate cross-domain search through links in LOD. The novelty of our method is that part of the LOD graph structure is regarded as a block image, and then image features of LOD are extracted. In experiments, we created about 30,000 literal pairs from a Japanese music category of DBpedia Japanese and Freebase, and confirmed that the proposed method correctly determines literal identity with F-measure of 99%.

Takahiro Kawamura, Shinichi Nagano, Akihiko Ohsuga

Audio Features in Music Information Retrieval

The rapid increase in the amount of on-line music services creates a need for automatic extraction of information from songs, music indexing and recommendation. But Music Information Retrieval is not an easy task and is still under development. This review paper focuses on features in Music Information Retrieval. We present audio engineering features and feature learning approach.

Daniel Grzywczak, Grzegorz Gwardys

Model of Auditory Filters and MPEG-7 Descriptors in Sound Recognition

It was examined whether applying a model of human auditory filter could improve the quality of sound recognition with the use of MPEG-7 standard audio descriptors. Modeling of filtering in the auditory system was with a bank of 38 gammatone filters closely spaced across the audible frequency range. The bank of filters was implemented as a low-level audio descriptor to replace the short-term Fourier transform (STFT) MPEG-7 audio descriptor. Sound recognition tests were conducted on a large set of sounds of nine musical instruments and speech of twelve speakers. The results showed that the proposed descriptor employing a bank of gammatone filters led to improved recognition of musical instruments and speakers as compared to the STFT-based original low-level MPEG-7 audio descriptor.

Aneta Świercz, Jan Żera

Security and Privacy in Ubiquitous Environment

A Scheme for Robust Biometric Watermarking in Web Databases for Ownership Proof with Identification

We propose a robust technique for watermarking relational databases using voice as a biometric identifier of ownership. The choice of voice as the distinguishing factor is influenced by its uniqueness, stability, universality, and ease of use. The watermark is generated by creating a statistical model of the features extracted from the owner’s voice. This biometric watermark is then securely embedded into selected positions of the fractional parts of selected numeric attributes using a reversible bit encoding technique. In case of a dispute regarding true ownership, the relative scores of the extracted watermark are generated by comparing features of the disputed voices with the extracted one. We experimentally demonstrate the robustness of the proposed technique against various attacks. Results show that watermark is extracted with 100% accuracy even when 97% tuples are added or when 90% tuples are deleted or when 80% tuples are altered. More significantly, biometric identification helped identify the correct owner even when only 60% of the watermark could be extracted.

Vidhi Khanduja, Shampa Chakraverty, Om Prakash Verma, Neha Singh

An Intelligent Multi-Agent Based Detection Framework for Classification of Android Malware

Android is currently the most popular operating system for smartphone devices with over 900 million installations until 2013. It is also the most vulnerable platform due to allowing of software downloads from 3rd party sites, loading additional code at runtime, and lack of frequent updates to known vulnerabilities. Securing such devices from malware that targets users is paramount. In this paper, we present a Jade agent based framework targeted towards protecting Android devices. We also focus on scenarios of use where such agents can be dynamically launched. We believe, a detection technique has to be intelligent due to limited battery constraints of these devices. Moreover, battery utilization might become secondary in certain settings where detection accuracy is given a higher preference. In this framework, the expensive analysis components utilizing machine-learning algorithms are pushed to server side, while agents on the Android client are used mainly for intelligent feature gathering.

Mohammed Alam, Son Thanh Vuong

Outlier Analysis Using Lattice of Contiguous Subspaces

Many anomaly detection techniques consider all the data-space dimensions when looking for outliers, and some others consider only specific subspaces, in isolation from other subspaces. However, interesting information about anomalous data points is embedded in the inter-relationships of the subspaces within which the data points appear to be outliers. Important characteristics of a dataset can be revealed by looking at these inter-relationships among subspaces. We describe a methodology for searching for outliers within the context of contiguous subspaces in the subspace lattice of a domain. We demonstrate additional insights about the outliers gained from this approach compared to finding the outliers in only specific subspaces or in the complete data-space. This additional information points an analyst to peculiar sets of subspaces to investigate further the underlying structure of the data space and also of the anomalous nature of the data points.

Vineet Joshi, Raj Bhatnagar

eSelect: Effective Subspace Selection for Detection of Anomalies

Anomaly detection is used for many applications such as detection of credit card fraud, medical diagnosis and computer system intrusion detection. Many interesting real-world data sets are high dimensional. Detection of anomalies in such datasets is hampered due to the curse of dimensionality. In such datasets the anomalies are hidden in smaller subspaces of attributes. However the number of subspaces possible from a given attribute set increases in a combinatorial fashion. Consequently an exhaustive search through all possible subspaces for anomalies is not computationally feasible. In this paper we propose a method for exploring the subspaces in a high dimensional data set in an effective and organized way to detect anomalies within them while avoiding an exhaustive search over all possible subspaces. The new method is called

eSelect

. Through extensive experimentation we compare

eSelect

to a well-established subspace selection method and demonstrate that our newly proposed method attains marked improvements.

Vineet Joshi, Raj Bhatnagar

Social Networks and Social Media

Active Recommendation of Tourist Attractions Based on Visitors Interests and Semantic Relatedness

Many visitors always search on tourist attractions related information on the Web so as to get more information on the places they are visiting or plan their next trips. In this study, we introduce CASIA-TAR, an active tourist attractions recommendation system, which provides relevant knowledge of specific tourist attractions and make recommendations for other relevant places to visit based on semantic relatedness among the specific tourist attraction and potentially interesting places. Two algorithms are introduced to calculate the semantic relatedness among different tourist attractions based on the tourist attraction semantic knowledge base with relevant knowledge mainly extracted from Web-based encyclopedias. As an integrated portal for tourist attraction recommendation, CASIA-TAR also provides images, news and microblog posts that are relevant to specific tourist attractions so that visitors could obtain relevant information in an integrated Web-based system.

Yi Zeng, Tielin Zhang, Hongwei Hao

Twitter Sentiment Polarity Analysis: A Novel Approach for Improving the Automated Labeling in a Text Corpora

The high penetration of Twitter in Chile has favored the use of this social network for companies and brands to get additional information on user opinions and feedback about their products and services. In recent years there have been many studies to determine the polarity of a comment on Twitter mainly considering three classes: positive, negative and neutral. One big difference inherent in the problem of Sentiment Analysis on Twitter as opposed to the Web is the ease with which you can obtain data to perform a supervised training algorithm with its API. To take advantage of this characteristic it is necessary to find a semi-automatic method for obtaining tweets to generate the corpora and avoid the traditional method of manual labeling which is very demanding in time and money. This paper goes deeper into the work of using a semi-automated generated corpora for Twitter sentiment polarity classification, introducing a novel approach of tweet selection for corpora consolidation and the addition of a fourth class of tweets that doesn’t correspond to any of the above. This new class includes tweets that are irrelevant for classification and do not contain much information, a type of posts that Twitter is full of. Experimental evaluations show that the usage of the fourth class of the denominated meaningless tweets together with the tweet filter criteria for corpus generation improves the system accuracy.

Pablo A. Tapia, Juan D. Velásquez

Empirical Study of Conversational Community Using Linguistic Expression and Profile Information

The popularization of social media exposes the structure of people’s conversation - what kind of people speak with whom, on what topics and with what kinds of words. In this paper, we propose a new approach to mining conversational network by community analysis, which exploits users’ profile information, interaction network and linguistic usage. Using our framework, we conducted empirical analysis on the complex relation among people’s profile information, social network, and language network using a large dataset from Twitter, which covers more than 7M people. Our findings include (1) we can extract a community composed of people who use the same kinds of slangs by exploiting information from both the social network and word usage, (2) when we focus on similarity among communities in terms of both interaction and word usage, we can find specific patterns based on the people’s profile information including their attributes and interests.

Junki Marui, Nozomi Nori, Takeshi Sakaki, Junichiro Mori

Multi Agent System Based Interface for Natural Disaster

Natural disasters cause devastation in the society due to unpredictable nature. Whether damage is minor or severe, emergency support should be provided within time. Multi-agent systems have been proposed to efficiently cope with emergency situations. Lot of work has been done on maturing core functionality of these systems but little attention has been given to their user interface. The world is moving towards an era where humans and machines work together to complete complex tasks. Management of such emergent situations is improved by combining superior human intelligence with efficiency of multi-agent systems. Our goal is to design and develop agents based interface that facilitates humans not only in operating the system but also in resource mobilization like ambulances, fire brigade, etc. to reduce life and property loss. This enhancement improves system adaptability and speeds up the relief operation by saving time of human-agent consumed in dealing with complex computer interface.

Zahra Sharmeen, Ana Maria Martinez-Enriquez, Muhammad Aslam, Afraz Zahra Syed, Talha Waheed

An Unsupervised Approach to Identify Location Based on the Content of User’s Tweet History

We propose and evaluate an unsupervised approach to identify the location of a user purely based on tweet history of that user. We combine the location references from tweets of a user with gazetteers like DBPedia to identify the geolocation of that user at a city level. This can be used for location based personalization services like targeted advertisements, recommendations and services on a finer level. In this paper, we use convex hull and k-center clustering, to identify the location of a user at a city level. The main contributions of this paper are: (i) reliability on just the contents of a tweet, without the need for manual intervention or training data; (ii) a novel approach to handle ambiguous location entries; and (iii) a computational geometric solution to narrow down the location of the user from a set of points corresponding to location references. Experimental results show that the system is able to identify a location for each user with high accuracy within a tolerance range. We also study the effect of tolerance on accuracy and average error distance.

Satya Katragadda, Miao Jin, Vijay Raghavan

Sensing Subjective Well-Being from Social Media

Subjective Well-being(

SWB

), which refers to how people experience the quality of their lives, is of great use to public policy-makers as well as economic, sociological research, etc. Traditionally, the measurement of SWB relies on time-consuming and costly self-report questionnaires. Nowadays, people are motivated to share their experiences and feelings on social media, so we propose to sense SWB from the vast user generated data on social media. By utilizing 1785 users’ social media data with SWB labels, we train machine learning models that are able to “sense” individual SWB. Our model, which attains the state-of-the-art prediction accuracy, can then be applied to identify large amount of social media users’ SWB in time with low cost.

Bibo Hao, Lin Li, Rui Gao, Ang Li, Tingshao Zhu

Special Sessions

Human Aspects in Cyber-Physical Systems

Detecting Stay Areas from a User’s Mobile Phone Data for Urban Computing

Nowadays, mobile phones are often used as an attractive option for large-scale sensing of human behavior, providing a source of real and reliable data for urban computing. As it is known to all, a user’s behavior often happened at some places where the user stayed over a certain time interval for a trip. For understanding a user’s behavior effectively, we need to detect the places where the user stayed over a certain time interval and we call these places stay areas. In this paper, we propose a method for detecting the stay areas from a user’s mobile phone data. The proposed method can tackle the complicated situations that the general method cannot deal with effectively. Through experimental evaluation, the proposed method is shown to deliver excellent performance.

Hui Wang, Ning Zhong, Zhisheng Huang, Jiajin Huang, Erzhong Zhou, Runqiang Du

Towards Robust Framework for On-line Human Activity Reporting Using Accelerometer Readings

This paper investigates subsequent matching approach and feature-based classification for activity recognition using accelerometer readings. Recognition is done by similarity measure based on Dynamic Time Warping (DTW) on each acceleration axis. Ensemble method is proposed and comparative study is executed showing better and more stable results. Our scenario assumes that activity is recognized with very small latency. Results shows that hybrid approach is promising for activity reporting, i.e. different walking patterns, using of tools. The proposed solution is designed to be a part of decision support in fire and rescue actions at the fire ground.

Michał Meina, Bartosz Celmer, Krzysztof Rykaczewski

Privacy-Preserving Emotion Detection for Crowd Management

Emotion detection plays a vital role in crowd management as it enables social event organizers to detect the actions of masses and react accordingly. There are several approaches to detect emotions in a crowd, including surveillance cameras, human observers and sensors. One other approach to gather emotion data is self-reporting. A recent study showed that self-reporting is feasible, reliable and efficient. However, there is a strong privacy concern among people that risks the use of such self-reporting mechanisms in wide use. In this work, we address the privacy aspect of self-reporting mechanism and propose a cryptographic approach that hides the sensitive data from the organizers but permits to compute statistical data for crowd management. The feasibility of using cryptography in real life for privacy protection is also investigated in terms of complexity.

Zeki Erkin, Jie Li, Arnold P. O. S. Vermeeren, Huib de Ridder

Cellular Automaton Evacuation Model Coupled with a Spatial Game

For web-based real-time safety analyses, we need computationally light simulation models. In this study, we develop an evacuation model, where the agents are equipped with simple decision-making abilities. As a starting point, a well-known cellular automaton (CA) evacuation model is used. In a CA, the agents move in a discrete square grid according to some transition probabilities. A recently introduced spatial game model is coupled to this CA. In the resulting model, the strategy choice of the agent determines his physical behavior in the CA. Thus, our model offers a game-theoretical interpretation to the agents’ movement in the CA.

Anton von Schantz, Harri Ehtamo

Facial Image Analysis with Applications

Shape from Motion Revisited

This brief tutorial paper on Shape from Motion (SfM), the profound 3D object modeling method, focuses on mathematical background for the batch scenario. Error bounds for pixels with respect to depth change are derived to analyze the applicability of orthographic projection versus perspective projection. Key geometric properties, used for SfM algorithms design and analysis, are stated and proved. Moreover, the case of measurement matrix with rank two, for non planar shapes is fully characterized and its role in shape ambiguity explored. Other sources of SfM ambiguity are presented what justifies the definition of ambiguous error function used for nonlinear optimization of rotation coefficients. Experiments refer to head pose identification and 3D face animation and show the good visual accuracy of SfM approach for digital 3D face projects.

Władysław Skarbek

Using Kinect for Facial Expression Recognition under Varying Poses and Illumination

Emotions analysis and recognition by the smartphones with front cameras is a relatively new concept. In this paper we present an algorithm that uses a low resolution 3D sensor for facial expression recognition. The 3D head pose as well as 3D location of the fiducial points are determined using Face Tracking SDK. Tens of the features are automatically selected from a pool determined by all possible line segments between such facial landmarks. We compared correctly classified ratios using features selected by AdaBoost, Lasso and histogram-based algorithms. We compared the classification accuracies obtained both on 3D maps and RGB images. Our results justify the feasibility of low accuracy 3D sensing devices for facial emotion recognition.

Filip Malawski, Bogdan Kwolek, Shinji Sako

High Accuracy Head Pose Tracking Survey

Head pose estimation is recently a more and more popular area of research. For the last three decades new approaches have constantly been developed, and steadily better accuracy was achieved. Unsurprisingly, a very broad range of methods was explored - statistical, geometrical and tracking-based to name a few. This paper presents a brief summary of the evolution of head pose estimation and a glimpse at the current state-of-the-art in this field.

Błażej Czupryński, Adam Strupczewski

Facial Expression Data Constructed with Kinect and Their Clustering Stability

In this paper, we construct facial expression benchmark data of 100 persons using Kinect face tracking application and study the stability of the benchmark data in terms of clustering. Kinect with its Software Development Kit applications has enabled low-cost constructions of various benchmark data on humans. We devised multi-lingual instruction sheets on 25 expressions, collected data from 115 persons, and carefully inspected and labeled the outcome to construct the data. The benchmark data consist of 263,106 instances, each of which includes 6 animation units, 11 shape units, and an image file all provided by the application. Out of the 263,106 instances, we labeled 62,500 of them as 1 of the 25 expressions and investigated their clustering stabilities to the 17 features. We show that the most frequently used clustering algorithm: k-means achieves the average normal mutual information about 0.92 as an evidence of the stability of our facial expression data.

Angdy Erna, Linli Yu, Kaikai Zhao, Wei Chen, Einoshin Suzuki

Eye-Gaze Tracking-Based Telepresence System for Videoconferencing

An approach to the teleimmersive videoconferencing system enhanced by the pan-tilt-zoom (PTZ) camera, controlled by the eye-gaze tracking system, is presented in this paper. An overview of the existing telepresence systems, especially dedicated to videoconferencing is included. The presented approach is based on the CyberEye eye-gaze tracking system engineered at the Multimedia Systems Department (MSD) of Gdańsk University of Technology (GUT), as well as on a standard PTZ security camera communicating with the computer by the TCP/IP protocol. Technical aspects of the developed system prototype including two different use cases (one-way and two-way configuration of system) are described. Moreover, a discussion related to the gathered user’s experience as well as to difficulties and opportunities concerning the proposed approach are included.

Bartosz Kunka, Adam Korzeniewski, Bożena Kostek, Andrzej Czyżewski

Workshops

Human Aspects in Ambient Intelligence

Using Process-Oriented Interfaces for Solving the Automation Paradox in Highly Automated Navy Vessels

This paper describes a coherent engineering method for developing high level human machine interaction within a highly automated environment consisting of sensors, actuators, automatic situation assessors and planning devices. Our approach combines ideas from cognitive work analysis, cognitive engineering, ontology engineering, and task-based prototyping. We describe our experiences with this approach when applying this suite to develop an innovative socio-technical system for fighting the internal battle in navy vessels with a strongly reduced manning.

Jurriaan van Diggelen, Wilfried Post, Marleen Rakhorst, Rinus Plasmeijer, Wessel van Staal

Multi-agent Solution for Adaptive Data Analysis in Sensor Networks at the Intelligent Hospital Ward

This paper introduces a multi-agent solution for remote monitoring based on wireless network of sensors that are used to collect and process medical data describing the current patient state. A multi-agent architecture is provided for a sensor network of medical devices, which is able to adaptively react to various events in real time. To implement this solution it is proposed to partially process the data by autonomous medical devices without transmitting it to the server and adapt the sampling intervals on the basis of the non-equidistant time series analysis. The solution is illustrated by simulation results and clinical deployment.

Anton Ivaschenko, Anton Minaev

The Challenges behind Independent Living Support Systems

Despite their crucial goal of assisting the elderly through their daily routine, Independent Living Support systems still are at their inception. This paper postulates that such systems be designed with a number of requirements in mind, and in particular with safety, security and privacy as fundamental ones. It then correspondingly articulates the three main challenges behind the development of Independent Living Support systems: requirement elicitation, design and correctness analysis. It is found that requirement elicitation will have to cope with a large variety of issues; that design will have to proceed from modularity; and, notably, that correctness analysis will have to be socio-technical. The last finding in particular emphasises that, for a system that prescribes vast interaction with the human, system correctness only makes sense if the system is analysed in combination with the human, rather than in isolation from the human. Building upon previous experience with the socio-technical analysis of Internet browsers, this paper identifies the specific socio-technical challenges that Independent Living Support systems pose, and indicates an approach to succeed in taking them.

Giampaolo Bella, Pekka Jäppinen, Jussi Laakkonen

Exploring Patterns as a Framework for Embedding Consent Mechanisms in Human-Agent Collectives

With ever increasing developments in computing technology, approaches to attaining informed consent are becoming outdated. In light of this ongoing change, researchers have begun to propose several new mechanisms to meet the emerging challenges of consent in pervasive settings. Unfortunately a particular problem arises when considering consent in the context of Human-Agent Collectives (HACs). These large-scale heterogeneous networks, of multiple co-operating humans and agents are particularly complex and it is difficult to know

what

,

where

and

how

to introduce these new mechanisms. In this paper we explore the potential of patterns of interactional arrangement as a framework for embedding consent mechanisms in HACs and other ubiquitous systems.

Stuart Moran, Ewa Luger, Tom Rodden

Increasing Physical and Social Activity through Virtual Coaching in an Ambient Environment

This paper describes the development and the validation of an ambient system (AAL-VU) that empowers its users in self-management of daily activities and social connectedness. The system combines state-of-the-art psychological knowledge on elderly user requirements for sustained behavior change with modern ICT technology in suggesting an adaptive ambient solution for the prevention and management of chronic diseases, inactivity and loneliness, thus resulting in a higher quality of life. Specifically, the AAL-VU system stimulates beneficial levels of activity in elderly as well as social connectedness. The focus on physical and social activity was chosen as this is recognized as a crucial element for the prevention, cure, and management of many chronic illnesses.

Arjen Brandenburgh, Ward van Breda, Wim van der Ham, Michel Klein, Lisette Moeskops, Peter Roelofsma

Enhancing Communication through Distributed Mixed Reality

A navigable mixed reality system where humans and agents can communicate and interact with each other in a virtual environment can be an appropriate tool for analyzing multi-human and multi-agent communication. We propose a prototype of our system, FCWorld, which has been developed to meet these requirements. FCWorld integrates various technologies with a focus on allowing natural human communication. In this paper we discuss the requirements for FCWorld, the technical issues which it must address, and our proposed solutions. We intend it to become a novel tool for a variety of communication tasks such as real-time analysis and facilitation.

Divesh Lala, Christian Nitschke, Toyoaki Nishida

A New Graphical Representation of the Chatterbots’ Knowledge Base

This paper discusses chatterbots that use the Pattern Recognition technique and the Artificial Intelligence Markup Language (AIML) regarding their design and implementation. Usually, chatterbot’s Knowledge Base (KB) is written with the AIML language, and it is based on the stimulus-response type block approach. Considering that the chatterbots’ KB might be a large collection of question-answer pairs, there is an increasing need of visual design proposals for the chatterbots’ dialogues. However, there is a lack of projects related to the chatterbots’ KB display. In order to fill this gap, this paper introduces a new graphical model called Dialogue Conceptual Diagram (DCD). It is a unified theoretical framework to describe the chatterbot’s knowledge in a graphical, intuitive and compact way. Finally, a subset of the ALICE’s KB was modeled using the DCDs, showing that DCD is a suitable framework to graphically represent the human-machine dialogues.

Maria das Graças Bruno Marietto, Rafael Varago de Aguiar, Wagner Tanaka Botelho, Robson dos Santos França

Sensing, Understanding, and Modeling for Smart City

Adaptive System for Intelligent Traffic Management in Smart Cities

In the paper we propose an adaptive system for intelligent traffic management in smart cities. We argue why the traffic is difficult and complex phenomenon, why such traffic management systems are necessary for the smart city, what is the state of the art in the traffic science and traffic management and how to improve existing solutions using methods that we develop, based on the Perception Based Computing paradigm.

Paweł Gora, Piotr Wasilewski

Implementing a Holistic Approach for the Smart City

Extending the services offered by the city requires, most of the times, reimplementation efforts. This paper presents our on-going efforts to develop a platform for the Smart City that focuses in providing the appropriate solutions for an easy integration of new services and devices. This endeavor is accomplished by abstracting communication issues using a middleware platform and by standardizing the way services are instantiating.

Roberto Requena, Antonio Agudo, Alba Baron, Maria Campos, Carlos Guijarro, Jose Puche, David Villa, Felix Villanueva, Juan Carlos Lopez

Elimination of Moving Shadow Based on Vibe and Chromaticity from Surveillance Videos

Shadow removal is one of the most important parts of moving object recognition in the field of intelligent video surveillance since the shadow definitely affects the recognition performance. This is caused from that shadows share the same movement patterns and similar magnitude of intensity to those of the foreground objects. Therefore, in this paper, to effectively remove moving shadows from video, a new approach based on chromaticity and a well-known universal background subtraction named as Vibe was proposed. Experimental results prove that moving shadows can be removed effectively by the proposed approach than the other ones.

Huaxiang Zhao, Min Wook Kang, Kyoung Yeon Kim, Yoo-Sung Kim

A Resemblance Based Approach for Recognition of Risks at a Fire Ground

This article focuses on a problem of a comparison between fire & rescue actions for a decision support at the fire ground. In our research, we split the actions into a set of frames which compose a timeline of a firefighting process. In our approach, the frames are represented as compound objects. We extract a set of features in order to represent these objects and we apply a comparator framework for the evaluation of similarities between the processes. The similarity constrains allow us to recognize the risks that appear during the actions. We justify our approach by showing results of a series of experiments which are based on reports describing real-life incidents.

Łukasz Sosnowski, Andrzej Pietruszka, Adam Krasuski, Andrzej Janusz

Backmatter

Weitere Informationen