• The Journal of the Acoustical Society of Korea
• /
• v.24 no.8
• /
• pp.498-504
• /
• 2005

We introduce a Korean speech recognition platform (ECHOS) developed for education and research Purposes. ECHOS lowers the entry barrier to speech recognition research and can be used as a reference engine by providing elementary speech recognition modules. It has an easy simple object-oriented architecture, implemented in the C++ language with the standard template library. The input of the ECHOS is digital speech data sampled at 8 or 16 kHz. Its output is the 1-best recognition result. N-best recognition results, and a word graph. The recognition engine is composed of MFCC/PLP feature extraction, HMM-based acoustic modeling, n-gram language modeling, finite state network (FSN)- and lexical tree-based search algorithms. It can handle various tasks from isolated word recognition to large vocabulary continuous speech recognition. We compare the performance of ECHOS and hidden Markov model toolkit (HTK) for validation. In an FSN-based task. ECHOS shows similar word accuracy while the recognition time is doubled because of object-oriented implementation. For a 8000-word continuous speech recognition task, using the lexical tree search algorithm different from the algorithm used in HTK, it increases the word error rate by $40\%$ relatively but reduces the recognition time to half.

• KIPS Transactions on Software and Data Engineering
• /
• v.7 no.5
• /
• pp.177-188
• /
• 2018

The scale of graph data has been increased rapidly because of the growth of mobile Internet applications and the proliferation of social network services. This brings upon the imminent necessity of efficient distributed and parallel graph processing approach since the size of these large-scale graphs are easily over a capacity of a single machine. Currently, there are two popular parallel graph processing approaches, vertex-centric graph processing and block centric processing. While a vertex-centric graph processing approach can easily be applied to the parallel processing system, a block-centric graph processing approach is proposed to compensate the drawbacks of the vertex-centric approach. In these systems, the initial quality of graph partition affects to the overall performance significantly. However, it is a very difficult problem to divide the graph into optimal states at the initial phase. Thus, several dynamic load balancing techniques have been studied that suggest the progressive partitioning during the graph processing time. In this paper, we present a load balancing algorithms for the block-centric graph processing approach where most of dynamic load balancing techniques are focused on vertex-centric systems. Our proposed algorithm focus on an improvement of the graph partition quality by dynamically reassigning blocks in runtime, and suggests block split strategy for escaping local optimum solution.

• Journal of IKEEE
• /
• v.5 no.1 s.8
• /
• pp.100-110
• /
• 2001

In this paper, we investigated several buffer management schemes for the design of shared-memory type ATM switches, which can enhance the utilization of switch resources and can support quality-of-service (QoS) functionalities. Our results show that dynamic threshold (DT) scheme demonstrate a moderate degree of robustness close to pushout(PO) scheme, which is known to be impractical in the perspective of hardware implementation, under various traffic conditions such as traffic loads, burstyness of incoming traffic, and load non-uniformity across output ports. Next, we considered buffer management strategies to support QoS functions, which utilize parameter values obtained via connection admission control (CAC) procedures to set tile threshold values. Through simulations, we showed that the buffer management schemes adopted behave well in the sense that they can protect regulated traffic from unregulated cell traffic in allocating buffer space. In particular, it was observed that dynamic partitioning is superior in terms of QoS support than virtual partitioning.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.37 no.6C
• /
• pp.476-482
• /
• 2012

In this paper, we analyze and compare several co-channel interference mitigation algorithms for WBAN application in 2.4 GHz ISM frequency bands. ML (Maximum Likelihood), OC (Optimal Combining) and MMSE (Minimum Mean Square Error) has been considered for the possible techniques for interference cancellation in view of the trade off between the performance and the complexity of implementation. Based on the channel model of IEEE 802.15.6 standard, simulation results show that ML and OC attains the lower BER performance than that of MMSE if we assume the perfect channel estimation. But, ML and OC have the additional requirement of implementation for his own and other users's channel estimation process, hence, besides the BER performance, the complexity of implementation and the sensitivity to channel estimation error should be considered since it requires the simple and small sized equipment for WBAN system application. In addition, the gap of detection BER performance between ML, OC and MMSE is much decreased under the imperfect channel estimation if we adopt real channel estimation process, therefore, in order to apply to WBAN system, the trade off between the BER performance and complexity of implemetation should be seriously considered to decide the best co-channel interference cancellation for WBAN system application.

• KIEE International Transaction on Systems and Control
• /
• v.2D no.2
• /
• pp.78-91
• /
• 2002

In the thermal power plant, there are six manipulated variables: main steam flow, feedwater flow, fuel flow, air flow, spray flow, and gas recirculation flow. There are five controlled variables: generator output, main steam pressure, main steam temperature, exhaust gas density, and reheater steam temperature. Therefore, the thermal power plant control system is a multinput and output system. In the control system, the main steam temperature is typically regulated by the fuel flow rate and the spray flow rate, and the reheater steam temperature is regulated by the gas recirculation flow rate. However, strict control of the steam temperature must be maintained to avoid thermal stress. Maintaining the steam temperature can be difficult due to heating value variation to the fuel source, time delay changes in the main steam temperature versus changes in fuel flow rate, difficulty of control of the main steam temperature control and the reheater steam temperature control system owing to the dynamic response characteristics of changes in steam temperature and the reheater steam temperature, and the fluctuation of inner fluid water and steam flow rates during the load-following operation. Up to the present time, the Proportional-Integral-Derivative Controller has been used to operate this system. However, it is very difficult to achieve an optimal PID gain with no experience, since the gain of the PID controller has to be manually tuned by trial and error. This paper focuses on the characteristic comparison of the PID controller and the modified 2-DOF PID Controller (Two-Degrees-Freedom Proportional-Integral-Derivative) on the DCS (Distributed Control System). The method is to design an optimal controller that can be operated on the thermal generating plant in Seoul, Korea. The modified 2-DOF PID controller is designed to enable parameters to fit into the thermal plant during disturbances. To attain an optimal control method, transfer function and operating data from start-up, running, and stop procedures of the thermal plant have been acquired. Through this research, the stable range of a 2-DOF parameter for only this system could be found for the start-up procedure and this parameter could be used for the tuning problem. Also, this paper addressed whether an intelligent tuning method based on immune network algorithms can be used effectively in tuning these controllers.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.48 no.5
• /
• pp.88-98
• /
• 2011

Recently, according as the importance of GIS(geography information system) database security is embossed, much researches had been achieved about GIS network security. But most such researches are weak against sourceful illegal reproductions and distributions of GIS vector data map. In this paper, we proposed an efficient layer unit contents based partial encryption technique in the vector map compression domain to prevent illegal distributions and unauthorized accesses. This method achieves a partial encryption about each central coordinate and directional parameters of a MCA(minimum coding attribute) that is created at the vector map compression processing in the vector space. First, the position encryption is applied as permutating randomly the center coordinate of each record that is minimum unit of vector map shape. And second, the direction encryption that changing shapes of vector map topography is applied as encrypting the direction of vertices's coordinates of each record. In experimental results, we confirmed that our proposed method can encipher the large volumed vector map data effectively in low computational complexity. Also, we could minimize the decline of compression efficiency that occurred by conventional contents based encryption schemes using AES or DES algorithms.

• Journal of Korea Water Resources Association
• /
• v.36 no.2
• /
• pp.195-209
• /
• 2003

Spatial-Stochastic Neural Networks Model(SSNNM) is used to estimate long-term streamflow in the parallel reservoir groups. SSNNM employs two kinds of backpropagation algorithms, based on LMBP and BFGS-QNBP separately. SSNNM has three layers, input, hidden, and output layer, in the structure and network configuration consists of 8-8-2 nodes one by one. Nodes in input layer are composed of streamflow, precipitation, pan evaporation, and temperature with the monthly average values collected from Andong and Imha reservoir. But some temporal differences apparently exist in their time series. For the SSNNM training procedure, the training sets in input layer are generated by the PARMA(1,1) stochastic model and they covers insufficient time series. Generated data series are used to train SSNNM and the model parameters, optimal connection weights and biases, are estimated during training procedure. They are applied to evaluate model validation using observed data sets. In this study, the new approaches give outstanding results by the comparison of statistical analysis and hydrographs in the model validation. SSNNM will help to manage and control water distribution and give basic data to develop long-term coupled operation system in parallel reservoir groups of the Upper Nakdong River.

• Proceedings of the Korean Institute of Navigation and Port Research Conference
• /
• v.1
• /
• pp.235-240
• /
• 2006

The paper will give an overview of the mission of GalTeC and then concentrate on two main aspects. The first more detailed aspect, is the analysis of the key performance parameters for the Galileo system services and presenting a technical overview of methods and algorithms used. The second more detailed aspect, is the service volume prediction including service dimensioning using the Prediction tool. In order to monitor and validate the Galileo SIS performance for Open Service (OS) and Safety Of Life services (SOL) regarding the key performance parameters, different analyses in the SIS domain and User domain are considered. In the SIS domain, the validation of Signal-in-Space Accuracy SISA and Signal-in-Space Monitoring Accuracy SISMA is performed. For this purpose first of all an independent OD&TS and Integrity determination and processing software is developed to generate the key reference performance parameters named as SISRE (Signal In Space Reference Errors) and related over-bounding statistical information SISRA (Signal In Space Reference Accuracy) based on raw measurements from independent sites (e.g. IGS), Galileo Ground Sensor Stations (GSS) or an own regional monitoring network. Secondly, the differences of orbits and satellite clock corrections between Galileo broadcast ephemeris and the precise reference ephemeris generated by GalTeC will also be compared to check the SIS accuracy. Thirdly, in the user domain, SIS based navigation solution PVT on reference sites using Galileo broadcast ephemeris and the precise ephemeris generated by GalTeC are also used to check key performance parameters. In order to demonstrate the GalTeC performance and the methods mentioned above, the paper presents an initial test result using GPS raw data and GPS broadcast ephemeris. In the tests, some Galileo typical performance parameters are used for GPS system. For example, the maximum URA for one day for one GPS satellite from GPS broadcast ephemeris is used as substitution of SISA to check GPS ephemeris accuracy. Using GalTeC OD&TS and GPS raw data from IGS reference sites, a 10 cm-level of precise orbit determination can be reached. Based on these precise GPS orbits from GalTeC, monitoring and validation of GPS performance can be achieved with a high confidence level. It can be concluded that one of the GalTeC missions is to provide the capability to assess Galileo and general GNSS performance and prediction methods based on a regional and global monitoring networks. Some capability, of which first results are shown in the paper, will be demonstrated further during the planned Galileo IOV phase, the Full Galileo constellation phase and for the different services particularly the Open Services and the Safety Of Life services based on the Galileo Integrity concept.

• The KIPS Transactions:PartD
• /
• v.11D no.2
• /
• pp.293-304
• /
• 2004

Recently, caching for the database-backed web applications has received much attention. The results of frequent queries could be cached for repeated reuse or for efficient processing of the relevant queries. Since the emergence of XML as a standard for data exchange on the web, today's web applications are to retrieve information from the remote XML sources across the network, and thus it is desirable to maintain the XML query results in the cache for the web applications. In this paper, we describe implementation of an XML query processing system that supports cache-answerability of XML queries, and evaluate its performance. XML path expression, which is one of the core features of XML query languages including XQuery, XPath, and XQL was considered as the XML query. Their result is maintained as an XML materialized view in the XML cache. The algorithms to rewrite the given XML path expression using its relevant materialized view proposed in [13] were implemented with RDBMS as XML store. The major issues of implementation are described in detail. The results of performance experiments conducted with the implemented system showed effectiveness of cache-answerability of XML queries. Comparison with previous research in terms of performance is also Provided.

• Journal of Internet Computing and Services
• /
• v.17 no.6
• /
• pp.53-59
• /
• 2016

Recently, huge stream data have been generated in real time from various applications such as wireless sensor networks, Internet of Things services, and social network services. For this reason, to develop an efficient method have become one of significant issues in order to discover useful information from such data by processing and analyzing them and employing the information for better decision making. Since stream data are generated continuously and rapidly, there is a need to deal with them through the minimum access. In addition, an appropriate method is required to analyze stream data in resource limited environments where fast processing with low power consumption is necessary. To address this issue, the sliding window model has been proposed and researched. Meanwhile, one of data mining techniques for finding meaningful information from huge data, pattern mining extracts such information in pattern forms. Frequency-based traditional pattern mining can process only binary databases and treats items in the databases with the same importance. As a result, frequent pattern mining has a disadvantage that cannot reflect characteristics of real databases although it has played an essential role in the data mining field. From this aspect, high utility pattern mining has suggested for discovering more meaningful information from non-binary databases with the consideration of the characteristics and relative importance of items. General high utility pattern mining methods for static databases, however, are not suitable for handling stream data. To address this issue, sliding window based high utility pattern mining has been proposed for finding significant information from stream data in resource limited environments by considering their characteristics and processing them efficiently. In this paper, we conduct various experiments with datasets for performance evaluation of sliding window based high utility pattern mining algorithms and analyze experimental results, through which we study their characteristics and direction of improvement.