Catálogo de publicaciones - libros

With more and more online game players having access to broadband Internet connections and high performance computers, peer-to-peer architecture offers an attractive solution for online multiplayer game design. However, message synchronization and cheat proof are two major challenges in implementing a fully distributed peer-to-peer game system. In this paper, a novel scheme, called , is proposed to support message synchronization and prevent protocol level cheats for such systems. In Semi-Lock, peers encrypt each message by using cryptographically secure one-way hash function. A two-step validation is applied at destination peers to verify the integrity and correctness of the received messages. Trace driven simulations are conducted to verify the performance of our proposed scheme. Simulation results show that it significantly outperforms existing algorithms in terms of message latency and message synchronization ratio, with only little extra computational overhead on each peer.

Recent advances in wireless communication and sensor and actuator technologies have played an essential role to realize the envisioned ubiquitous world. Wireless sensor networks (WSNs) have emerged as a great catalyst for morphing personal area networks (PANs) into low power personal area networks (LoWPANs) which itself is a channel to achieve higher degrees of ubiquity and pervasiveness. These LoWPANs need to be connected with other wireless and wired networks in order to maximize the utilization of information and other resources which are mainly associated with the IP networks. Interworking of LoWPANs with IP networks brings in many challenges for service discovery and network selection. A great problem in this scenario is to find and use services in the closest proximity of the user. In this paper we propose novel service discovery architecture and algorithms that help proximity based service discovery and network selection within an IP network and LoWPAN interworked environment. The results show that our architecture helps finding and using the closest services from inside as well as outside the LoWPAN. It also reduces the traffic overhead for service discovery considerably as compared to other protocols.

In the ubiquitous computing environment, wearable computers are becoming more and more important because they provide an interface between human and computers. However, there are many challenges that come from its distributed, dynamic, heterogeneous computing environment. Most of all, an wearable computer has limited I/O devices for user interface. Thus, there are many context-aware applications to provide convenience for users. But each application should suffer from finding corresponding data. So, it is very important to support context-awareness in a file service for these applications through maintaining user’s activities and metadata derived from them.

In this paper, we present a context-aware file service suitable for the wearable computing environment to provide a convenient data sharing service among devices and users. To realize our work, we present an ontology based file metadata management scheme for context awareness.

Finally, we implement the early prototype by using the UPnP[7] and WebDAV[8] technologies and define source ontology by OWL[9] and are still refining the system design. We believe that this work could be a predominant reference for a new generation of file management technology in the ubiquitous computing environment.

Many of the systems developed to date for controlling home appliances remotely from outside have used a web server as the interface between the home network and the users. Although the user interface (UI) that uses a web server offers many advantages, this approach requires a fixed IP (Internet Protocol) address and does not have a push function. Moreover, until the user gets reconnected to the web server, he or she cannot know the result of a submitted order. In particular, internet-based applications using web browsers still require considerable time to execute the initialization process for acquiring state information on home appliances from the home server. To solve these problems, we propose an application for home automation that can efficiently control and monitor home appliances using an instant messaging system (IMS) with real time communication. The proposed system has not only the advantages of the web server method, but also with additional advantage that the homeowner does not need to continually reconnect to the home server, and that the home server does not need to have a fixed IP address. Our application can be applied using middleware technologies.

Context delivery is an inevitable issue for ubiquitous computing. Context-aware middlewares perform all the functions of context sensing, inferring and delivery to context-aware applications. But one of the major issues for these middlewares is to devise a context delivery scheme that is scalable as well as efficient. Pure unicast or pure broadcast based dissemination can not provide scalability as well as less average latency. In this paper we present a scalable context delivery mechanism for context-aware middlewares based on hybrid data dissemination technique where the most requested data are broadcasted and the rest are delivered through unicast. Our scheme is adaptive in the sense that it dynamically differentiates hot (most requested) and cold (less requested) data according to request rate and waiting time. Inclusion of lease mechanism and bandwidth division further allows us to reduce network traffic and average latency. We validated our claim through extensive simulation.

For an agent to be believable, emotion is an essential part of the human-computer interaction. In this paper, we propose a dynamic affective system, inspired from neuroscience. The system comprises four modules: an appraisal module, an emotion-generation module, an emotional expression module, and a long-term memory (LTM). The appraisal module gathers environmental stimuli and evaluates them to see if they are rewarding or punishing. Based on the results of the appraisal module and the agent’s internal stimuli, the emotion-generation module generates the affective states of the agent. The emotional expression module combines and expresses emotional behaviors in accordance with the affective states. As a result, the proposed affective system can generate various emotions simultaneously and produces emotional life-like expressions.

In this paper, we propose a novel face recognition method using the effective transform coefficients of face images. The method is based on extraction of effective vector in Discrete Cosine Transform (DCT) domain and Linear Discriminant Analysis (LDA) of effective vector. In general, face images have characteristics that they show larger energy congestion in horizontal frequency coefficients than in vertical or diagonal frequency coefficients. However, many previous methods have shortcomings that they don’t utilize the facial energy characteristics. To overcome shortcomings above, the proposed method selects the effective coefficients of the face in DCT domain and then extracts feature vector through LDA analysis on DCT coefficients. Experimental results show that our method has improvements of recognition performance over the previous methods.

In this paper we describe our approach to find the most suitable among several projection based Augmented Reality systems to display information on moveable objects in real time. After an introduction to the scenario, we describe our approach to estimate the Quality of Projection that each individual projection system can achieve. After that we describe how we compared these estimates to the perception of human beings. Finally we present the results, which includes a new estimation function.

Context-aware computing is a hot area in ubiquitous computing. There are several challenges to be covered. This paper focuses on context reasoning, which means deducing higher context from raw sensor data. The context reasoning problem is discussed on two different levels: context inference/recognition concerning the generation of context atoms from raw sensor data, and context reasoning concerning the composition of context atoms and deduction of higher-level context. In this paper, we discuss some commonly used reasoning technologies in context-aware systems, including rule-based logics and machine learning methods. Besides, a clustering algorithm, the Symbol Clustering Map, is introduced to learn the current context.

As web documents proliferate fast, the need fo real-time computation of change (edit script) between web documents increases. Though fast heuristic algorithms have been proposed recently, the qualities of edit scripts produced by them are not satisfactory. In this paper, we propose X-tree Diff+ which produces better quality of edit scripts by introducing a tuning step based on the notion of consistency of matching. We also add copy operation to provide users more convenience. Tuning and copy operation increase matching ratio drastically. X-tree Diff+ produces better quality of edit scripts and runs fast equivalent to the time complexity of fastest heuristic algorithms.