|A HMM-Based System for Training of Second Language Aquisition|
Lingyun Gu, John G. Harris
We describe a system for the training of Second Language Acquisition Pronunciation (SLAP) for non-native speakers. This speech recognition-based system is designed to mimic the valuable interactions between second-language students and a fluent teacher. When a student speaks a word into SLAPís microphone, it is analyzed to determine the part of the word (if any) that is incorrectly pronounced. A fluent utterance of the word is then played back to the student with emphasis on the mispronounced part of the word. Just as a live teacher naturally does, the difficult part of the word is played back louder, extended in time and possibly with higher pitch. We demonstrate SLAP on a multisyllabic word to show typical performance.
A Process For Performance Evaluation Of Real-Time Systems
Andrew J. Kornecki, Eric Sorton
Real-time developers and engineers must not only meet the system functional requirements, but also the stringent timing requirements. One of the critical decisions leading to meeting these timing requirements is the selection of an operating system under which the software will be developed and run. Although there is ample documentation on real-time systems performance and evaluation, little can be found that combines such information into an efficient process for use by developers. As the software industry moves towards clearly defined processes, creation of appropriate guidelines describing a process for performance evaluation of real-time system would greatly benefit real-time developers. This technology transition research focuses on developing such a process. PROPERT (PROcess for Performance Evaluation of Real Time systems) - the process described in this paper - is based upon established techniques for evaluating real-time systems. It organizes already existing real-time performance criteria and assessment techniques in a manner consistent with a well-formed process, based on the Personal Software Process concepts.
A Synchronisation Method For Informed Spread-Spectrum Audiowatermarking
Pierre-Yves Fulchiron, Barry O'Donovan, Guenole Silvestre, Neil Hurley
Under perfect synchronisation conditions, watermarking schemes employing asymmetric spread-spectrum techniques are suitable for copy-protection of audio signals. This paper proposes to combine the use of a robust psychoacoustic projection for the extraction of a watermark feature vector along with non-linear detection functions optimised with side-information. The new proposed scheme benefits from an increased level of security through the use of asymmetric detectors. We apply this scheme to real audio signals and experimental results show an increased robustness to desynchronisation attacks such as random cropping.
A Synthesized Framework for Formal Verification of Computing Systems
Nikola Bogunovic, Igor Grudenic, Edgar Pek
Design process of computing systems gradually evolved to a level that encompasses formal verification techniques. However, the integration of formal verification techniques into a methodical design procedure has many inherent miscomprehensions and problems. The paper explicates the discrepancy between the real system implementation and the abstracted model that is actually used in the formal verification procedure. Particular attention is paid to the seamless integration of all phases of the verification procedure that encompasses definition of the specification language and denotation and execution of conformance relation between the abstracted model and its intended behavior. The concealed obstacles are exposed, computationally expensive steps identified and possible improvements proposed.
A Three-layered Self-Organizing Map Neural Network for Clustering Analysis
Sheng-Chai Chi, Chi-Chung Lee, Tung-Chang Young
In the commercial world today, holding the effective information through information technology (IT) and the internet is a very important indicator of whether an enterprise has competitive advantage in business. Clustering analysis, a technique for data mining or data analysis in databases, has been widely applied in various areas. Its purpose is to segment the individuals in the same population according to their characteristics. In this research, an enhanced three-layered self-organizing map neural network, called 3LSOM, is developed to overcome the drawback of the conventional two-layered SOM through sight-inspection after the mapping process. To further verify its feasibility, the proposed model is applied to two common problems: the identification of four given groups of work-part images and the clustering of a machine/part incidence matrix. The experimental results prove that the data that belong to the same group can be mapped to the same neuron on the output layer of the 3LSOM. Its performance in clustering accuracy is good and is also comparable with that of the FSOM, FCM and k-Means.
A Virtual Environments Editor for Driving Scenes
Ronald R. Mourant, Sophia-Katerina Marangos
The goal of this project was to enable the rapid creation of three-dimensional virtual driving environments. We designed and implemented a high-level scene editor that allows a user to construct a driving environment by pasting icons that represent 1) road segments, 2) road signs, 3) trees and 4) buildings. These icons represent two- and three-dimensional objects that have been predesigned. Icons can be placed in the scene at specific locations (x, y, and z coordinates). The editor includes the capability of a user to "drive" a vehicle using a computer mouse for steering, accelerating and braking. At any time during the process of building a virtual environment, a user may switch to "Run Mode" and inspect the three-dimensional scene by "driving" through it using the mouse. Adjustments and additions can be made to the virtual environment by going back to "Build Mode". Once a user is satisfied with the threedimensional virtual environment, it can be saved in a file. The file can used with Java3D software that enables the traversing of three-dimensional environments. The process of building virtual environments from predesigned icons can be applied to many other application areas. It will enable novice computer users to rapidly construct and use three-dimensional virtual environments.
Adaptive Queue Management with Restraint on Non-Responsive Flows
Lan Li, Gyungho Lee
This paper proposes an adaptive queue management scheme (adaptive RED) to improve Random Early Detection (RED) on restraining non-responsive flows. Due to a lack of flow control mechanism, non-responsive flows can starve responsive flows for buffer and bandwidth at the gateway. In order to solve the disproportionate resource problem, RED framework is modified in this way: on detecting when the non-responsive flows starve the queue, packet-drop intensity (Max_p in RED) can be adaptively adjusted to curb non-responsive flows for resource fair-sharing, such as buffer and bandwidth fair-sharing. Based on detection of traffic behaviors, intentionally restraining nonresponsive flows is to increase the throughput and decrease the drop rate of responsive flows. Our experimental results based on adaptive RED shows that the enhancement of responsive traffic and the better sharing of buffer and bandwidth can be achieved under a variety of traffic scenarios.
An Efficient Scheme for Aggregation and Presentation of Network Performance in Distributed Brokering Systems
Gurhan Gunduz, Shrideep Pallickara, Geoffrey Fox
The Internet is presently being used to support increasingly complex interaction models as a result of more and more applications, services and frameworks becoming network centric. Efficient utilization of network and networkedresources is of paramount importance. Network performance gathering is a precursor to any scheme that seeks to provide adaptive routing capabilities for interactions. In this paper we present a network performance aggregation framework that is extensible and appropriate for distributed messaging systems that span multiple realms, disparate communication protocols and support different applications.
Building RDF-Schema Based Knowledge Map for Knowledge Management Systems
Doug W. Choi, Hyun Gyu Lee
This paper attempts to implement the semantic web paradigm, which enables the semantic search of knowledge and information on the web environment, by adopting RDF, RDF-schema and Dublin core data element structure in the process of building knowledge map. Since the RDF, RDF-schema, and Dublin core data element structure comply with the XML grammar, we have greater convenience and flexibility in representing knowledge into XML document format.
Representing knowledge following the RDF-schema and the Dublin core structure enables the well-formed and semantic description of knowledge, facilitating the representation of knowledge into the frame format.
|Delegation using A Proxy Certificate in On-line|
Taesung Kim, Sangrae Cho, Seunghun Jin
Delegation is frequently used in real world. In order to make it possible in on-line, the fact of delegation should be protected from malicious modification and the right of a proxy should be controlled properly. This paper describes security requirements and proxy certificate and proposes a practical method of restriction of a proxy. Finally, the prototype implementation is addressed.
Digital Watermarks Using Discrete Wavelet Transformation and Spectrum Spreading
Ryousuke Takai, Kenji Nagasaka
In recent tears, digital media makes rapid progress through the development of digital technology. Digital media normally assures fairly high quality, nevertheless can be easily reproduced in a perfect form. This perfect reproducibility takes and advantage from a certain point of view, while it produces an essential disadvantage, since digital media is frequently copied illegally. Thus the problem of the copyright protection becomes a very important issue. A solution of this problem is to embed digital watermarks that is not perceived clearly by usual people, but represents the proper right of original product.
In our method, the images data in the frequency domain are transformed by the Discrete Wavelet Transform and analyzed by the multi resolution approximation, . Further, the spectrum spreading is executed by using PN-sequences.
Choi and Aizawa  embed watermarks by using block correlation of DCT coefficients. Thus, we apply Discrete Cosine Transformation, abbreviated to DCT, instead of the Fourier transformation in order to embed watermarks.
If the value of this variance is high then we decide that the block has bigger magnitude for visual fluctuations. Henceforth, we may embed stronger watermarks, which gives resistance for images processing, such as attacks and/or compressions.
Efficient Pruning Method for Ensemble Self-Generating Neural Networks
Hirotaka Inoue, Hiroyuki Narihisa
Recently, multiple classifier systems (MCS) have been used for practical applications to improve classification accuracy. Self-generating neural networks (SGNN) are one of the suitable base-classifiers for MCS because of their simple setting and fast learning. However, the computation cost of the MCS increases in proportion to the number of SGNN. In this paper, we propose an efficient pruning method for the structure of the SGNN in the MCS. We compare the pruned MCS with two sampling methods. Experiments have been conducted to compare the pruned MCS with an unpruned MCS, the MCS based on C4.5, and k-nearest neighbor method. The results show that the pruned MCS can improve its classification accuracy as well as reducing the computation cost.
Information System Architectures: Representation, Planning and Evaluation
Andrť Vasconcelos, Pedro Sousa, Josť Tribolet
In recent years organizations have been faced with increasingly demanding business environments - pushed by factors like market globalization, need for product and service innovation and product life cycle reduction - and with new information technologies changes and opportunities- such as the Component-off-the-shelf paradigm, the telecommunications improvement or the Enterprise Systems off-the-shelf modules availability - all of which impose a continuous redraw and reorganization of business strategies and processes.
Nowadays, Information Technology makes possible high-speed, efficient and low cost access to the enterprise information, providing the means for business processes automation and improvement. In spite of these important technological progresses, information systems that support business, do not usually answer efficiently enough to the continuous demands that organizations are faced with, causing non-alignment between business and information technologies (IT) and therefore reducing organization competitive abilities.
This article discusses the vital role that the definition of an Information System Architecture (ISA) has in the development of Enterprise Information Systems that are capable of staying fully aligned with organization strategy and business needs. In this article the authors propose a restricted collection of founding and basis operations, which will provide the conceptual paradigm and tools for proper ISA handling. These tools are then used in order to represent, plan and evaluate an ISA of a Financial Group.
Inverse Problem In Optical Tomography Using Diffusion Approximation and Its Hopf-Cole Transformation
Taufiquar R. Khan
In this paper, we derive the Hopf-Cole transformation to the diffusion approximation. We find the analytic solution to the one dimensional diffusion approximation and its Hopf-Cole transformation for a homogenous constant background medium. We demonstrate that for a homogenous constant background medium in one dimension, the Hopf-Cole transformation improves the stability of the inverse problem. We also derive a Green's function scaling of the higher dimensional diffusion approximation for an inhomogeneous background medium and discuss a two step reconstruction algorithm.
Selection of the Most Indicative Wavelets for the Multiresolution of the Vowels
Ismet Traljic, B. Perunicic, Melita Ahic-Dokic
The paper describes one approach of the selection of the most indicative wavelets for each of the vowels in the authorís native language. Analysis is performed on the correct and incorrect vowels. On each of the sample multiresolution decomposition is applied. For each of the detail and approximation the most indicative wavelet is selected using value of the variance as the criteria. Some interesting results are obtained and biorthogonal wavelets have been select as the most appropriate for the multiresolution of the vowels. Using this criterion, any further analysis of the samples can be done using only coefficients of the discrete wavelet transformation on the level of approximation or any level of the detail, with enough guarantees that they are most appropriate for each vowel.
Systematic Approach for Scheduling of Tasks and Messages under Noise Environment
Hyoung Yuk Kim, Hye Min Shin, Hong Seong Park
High degree of EMI (Electro-Magnetic Interference) or noise is generated in the plant environment, where control systems consist of smart sensors, smart actuators, and controllers connected via a fieldbus. The noise generated by some devices such as high-power motors may cause communication errors and delay the successful transmission of data. Therefore, the noise condition is one of the important parameters considered in the design of a reliable network-based control system. This paper presents the scheduling method of task and message to guarantee the given end-to-end constraints under noise environments. The presented scheduling method is applied to an example of a control system that uses CAN (Controller Area Network), considering two kinds of noise models. The comparison results for each noise condition shows the importance of considering the noise condition in system design. System designers are able to design the control system, guaranteeing its requirements under a noise environment by using the proposed scheduling method.
Ultrafast Hierarchical OTDM/WDM Network
Hideyuki Sotobayashi, Wataru Chujo, Takeshi Ozeki
Ultrafast hierarchical OTDM/WDM network is proposed for the future core-network. We review its enabling technologies: C- and L-wavelength-band generation, OTDM-WDM mutual multiplexing format conversions, and ultrafast OTDM wavelengthband conversions.