Catálogo de publicaciones - libros

Recently there have been a large number of multimedia demands of IP-based applications in tremendous Internet world. However, Internet-based multimedia services typically do not offer any quality of service because of its own characteristics such as network bandwidth. In case of Korea, assumption that dense residential communities sharing their livings and information have their infrastructures like high speed communication network is able to achieve quality of service at a close range. This trend creates new market by creating new type of service in Korean-style Internet server. In this paper, we design and implement the high performance media streaming server focusing on local communities which obtain high speed network bandwidth, using streaming-accelerated network storage card which is enable to offer ”zero-copy” interfaces and raw-disk I/O.

An advanced multimedia processing unit is needed for the DVD player, the music player, and the video game machine, etc., we propose a sub-processor for advanced multimedia processing for built-in usage. We achieve the necessary arithmetic capacity equal with the processing, only for the multimedia by proposing multimedia processing is done by extended instruction. The tile processor is adopted as architecture of the processor. The tile processor can reduce the wiring delay that happens because of tile positioning, and is suitable for the miniaturization required for built-in processors. Moreover, the high parallel processing ability of the tile architecture is the best for media operations involving same operation and repetitions of data volume. With the aim of time efficiency, SIMD style operation was installed. However with device miniaturization a high level of integrations necessary

Presently, the wireless mobile networks and devices are becoming increasingly popular to provide users the access anytime and anywhere. The mobile systems are based on cellular approach and the area is covered by cells that overlap each other. In mobile cellular systems the handover is a very important process, which refers to a mechanism that transfers an ongoing call from one Base Station (BS) to another. The performance of the handover mechanism is very important to maintain the desired Quality of Service (QoS). Many handover algorithms are proposed in the literature. However, to make a better handover and keep the QoS in wireless networks is very difficult. In this paper, we propose a fuzzy-based speed-aware handover system. The performance evaluation via simulations shows that proposed system has a good handover decision.

The traditional fusion methods of multiple sensing modalities are summarized with 1) data-level fusion, 2) feature-level fusion and 3) decision-level fusion. This paper suggests the decision-level fusion-oriented novel fusion and fission framework, and it implements WPS (Wearable Personal Station) and Voice-XML-Based Multi-Modal Fusion Agent (hereinafter, MMFA) using audio-gesture modalities. Because the MMFA provides different weight and a feed-back function in individual recognizer, according to SNNR(Signal Plus Noise to Noise Ratio) and fuzzy value, it may select an optimal instruction processing interface under a given situation or noisy environment, and can allow more interactive communication functions in noisy environment. In addition, the MMFA provides a wider range of personalized information more effectively as well as it not need complicated mathematical algorithm and computation costs that are concerned with multidimensional features and patterns (data) size, according as it use a WPS and distributed computing-based database and SQL-logic, for synchronization and fusion between modalities.

Grid computing is an emerging computing architecture that can solve massive computational problems by making use of large numbers of heterogeneous computers. Job scheduling is an important issue in the high performance Grid computing environment. An appropriate scheduling algorithm can efficiently reduce the response time, turnaround time and further increase the throughput. However, finding an optimal grid scheduling algorithm is intractable. In this paper, we propose a high performance scheduling algorithm based on Fuzzy Neural Networks to resolve this problem. In the proposed algorithm, we apply the Fuzzy Logic technique to evaluate the grid system load status, and adopt the Neural Networks to automatically tune the membership functions. Since there are many factors that influence the system’s load circumstances; as the number of factors increase, it becomes very difficult to set up the system using general experience. We implemented a Fuzzy Neural Network scheduler based on Globus Toolkit 4 to verify the proposed scheduling algorithm performance. NAS Grid Benchmarks (NGB) was utilized to validate the performance of our scheduling approach. The experimental results show that our proposed algorithm can reduce the turnaround time and has better speed-up ratio than previous methods.