• 대한전기학회논문지
• /
• 제37권9호
• /
• pp.663-670
• /
• 1988

Advances in VLSI technology have made it possible to have a larger number of processing elements to be included in highly parallel processor system. A system with a large number of processing elements and memory requires a complex data path. Multistage Interconnection networks(MINS) are useful in providing programmable data path between processing elements and memory modules in multiprocessor system. In this thesis, the performance of MINS for the star network has been analyzed and compared with other networks, such as generalized shuffle network, delta network, and referenced crossbar network.

• Journal of Biosystems Engineering
• /
• 제25권5호
• /
• pp.409-414
• /
• 2000

A machine vision system was built to obtain single stationary image from an egg. This system includes a CCD camera, an image processing board and a lighting system. A computer program was written to acquire, enhance and get histogram from an image. To minimize the evaluation time, the artificial neural network with the histogram of the image was used for eggshell evaluation. Various artificial neural networks with different parameters were trained and tested. The best network(64-50-1 and 128-10-1) showed an accuracy of 87.5% in evaluating eggshell. The comparison test for the elapsed processing time per an egg spent by this method(image processing and artificial neural network) and by the processing time per an egg spent by this method(image processing and artificial neural network) and by the previous method(image processing only) revealed that it was reduced to about a half(5.5s from 10.6s) in case of cracked eggs and was reduced to about one-fifth(5.5s from 21.1s) in case of normal eggs. This indicates that a fast eggshell evaluation system can be developed by using machine vision and artificial neural network.

• 대한원격탐사학회:학술대회논문집
• /
• 대한원격탐사학회 2007년도 Proceedings of ISRS 2007
• /
• pp.101-104
• /
• 2007

Efficient Query processing and optimization are critical for reducing network traffic and decreasing latency of query when accessing and manipulating sensor data of large-scale sensor networks. Currently it has been studied in sensor database projects. These works have mainly focused on in-network query processing for sensor networks and assumes homogeneous sensor networks, where each sensor network has same hardware and software configuration. In this paper, we present a framework for efficient query processing over heterogeneous sensor networks. Our proposed framework introduces query processing paradigm considering two heterogeneous characteristics of sensor networks: (1) data dissemination approach such as push, pull, and hybrid; (2) query processing capability of sensor networks if they may support in-network aggregation, spatial, periodic and conditional operators. Additionally, we propose multi-query optimization strategies supporting cross-translation between data acquisition query and data stream query to minimize total cost of multiple queries. It has been implemented in WSN middleware, COSMOS, developed by ETRI.

• 한국음향학회:학술대회논문집
• /
• 한국음향학회 1995년도 제12회 음성통신 및 신호처리 워크샵 논문집 (SCAS 12권 1호)
• /
• pp.33-41
• /
• 1995

We proposed the approach method of neural network for signal processing, especially speech signal processing and reviewed the algorithms for several neural networks which are used for many alppication field in speech processing. Finally, investigated the trends in neural network method through 3 conference jounal and the ASK jounal in 1994.

• 한국정보과학회논문지:데이타베이스
• /
• 제29권6호
• /
• pp.492-501
• /
• 2002

비동기 트리거 처리기(ATP)는 데이타베이스 트랜잭션의 수행이 완료된 후에 트리거를 처리하는 소프트웨어 시스템이다. ATP 내에서는 트리거 조건의 효율적인 검사를 위하여 차별화 네트워크(discrimination network)가 사용된다. 차별화 네트워크는 내부 상태를 메모리 노드에 저장한다. TrigerMan은 하나의 ATP로써 차별화 네트워크로써 Gator 네트워크를 사용한다. 데이타베이스의 내용 변화는 트리거맨에 토큰 형태로 전달된다. 트리거 조건의 검사는 토큰이 Gator 네트워크를 통과하면서 이루어지는데, 이때 Gator 네트워크의 메모리 노드들이 갱신된다. 토큰의 병렬처리는 시스템의 성능을 향상시키는 여러 방법 중 하나이지만 통제되지 않은 병렬처리는 잘못된 트리거 액션 수행을 유발한다. 이 논문은, 최소한의 이상 현상만을 허용하며 토큰의 병렬 처리를 가능하게 하는, 네 가지 트리거 처리 일관성 레벨을 제안한다. 우리는 각 일관성 레벨에 대하여 병렬 토큰 처리를 가능하게 하는 고유한 기술을 개발하였다. 제안된 기술은 안정된 방법이라는 사실이 공리를 통하여 증명되었으며, 이 기술은 실체화 된 (materialized) 뷰 유지 (view maintenance)에 사용될 수 있다.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2008년도 추계학술발표대회
• /
• pp.874-877
• /
• 2008

Nowadays the existing data processing systems can only support some simple query for sensor network. It is increasingly important to process the vast data streams in sensor network, and achieve effective acknowledges for users. In this paper, we propose a holistic distributed k-means algorithm for sensor network. In order to verify the effectiveness of this method, we compare it with central k-means algorithm to process the data streams in sensor network. From the evaluation experiments, we can verify that the proposed algorithm is highly capable of processing vast data stream with less computation time. This algorithm prefers to cluster the data streams at the distributed nodes, and therefore it largely reduces redundant data communications compared to the central processing algorithm.

• Journal of Information Processing Systems
• /
• 제14권6호
• /
• pp.1464-1479
• /
• 2018

In face recognition, sometimes the number of available training samples for single category is insufficient. Therefore, the performances of models trained by convolutional neural network are not ideal. The small sample face recognition algorithm based on novel Siamese network is proposed in this paper, which doesn't need rich samples for training. The algorithm designs and realizes a new Siamese network model, SiameseFacel, which uses pairs of face images as inputs and maps them to target space so that the $L_2$ norm distance in target space can represent the semantic distance in input space. The mapping is represented by the neural network in supervised learning. Moreover, a more lightweight Siamese network model, SiameseFace2, is designed to reduce the network parameters without losing accuracy. We also present a new method to generate training data and expand the number of training samples for single category in AR and labeled faces in the wild (LFW) datasets, which improves the recognition accuracy of the models. Four loss functions are adopted to carry out experiments on AR and LFW datasets. The results show that the contrastive loss function combined with new Siamese network model in this paper can effectively improve the accuracy of face recognition.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제6권7호
• /
• pp.1756-1776
• /
• 2012

Wireless sensor networks (WSNs) are likely to be more prevalent as their cost-effectiveness improves. The spectrum of applications for WSNs spans multiple domains. In environmental sciences, in particular, they are on the way to become an essential technology for monitoring the natural environment and the dynamic behavior of transient physical phenomena over space. Existing sensor network query processors (SNQPs) have also demonstrated that in-network processing is an effective and efficient means of interaction with WSNs for performing queries over live data. Inspired by these findings, this paper investigates the question as to whether spatio-temporal and historical analysis can be carried over WSNs using distributed query-processing techniques. The emphasis of this work is on the spatial, temporal and historical aspects of sensed data, which are not adequately addressed in existing SNQPs. This paper surveys the novel approaches of storing the data and execution of spatio-temporal and historical queries. We introduce the challenges and opportunities of research in the field of in-network storage and in-network spatio-temporal query processing as well as illustrate the current status of research in this field. We also present new areas where the spatio-temporal and historical query processing can be of significant importance.

• 한국항해항만학회:학술대회논문집
• /
• 한국항해항만학회 2006년도 International Symposium on GPS/GNSS Vol.1
• /
• pp.171-174
• /
• 2006

A large number of service providers in countries all over the world have established GNSS reference station networks in the last years and are using network software today to provide a correction stream to the user as a routine service. In current GNSS network processing, all the geometric related information such as ionospheric free carrier phase ambiguities from all stations and satellites, tropospheric effects, orbit errors, receiver and satellite clock errors are estimated in one centralized Kalman filter. Although this approach provides an optimal solution to the estimation problem, however, the processing time increases cubically with the number of reference stations in the network. Until now one single Personal Computer with Pentium 3.06 GHz CPU can only process data from a network consisting of no more than 50 stations in real time. In order to process data for larger networks in real time and to lower the computational load, a federated filter approach can be considered. The main benefit of this approach is that each local filter runs with reduced number of states and the computation time for the whole system increases only linearly with the number of local sensors, thus significantly reduces the computational load compared to the centralized filter approach. This paper presents the technical aspect and performance analysis of the federated filter approach. Test results show that for a network of 100 reference stations, with the centralized approach, the network processing including ionospheric modeling and network ambiguity fixing needs approximately 60 hours to process 24 hours network data in a 3.06 GHz computer, which means it is impossible to run this network in real time. With the federated filter approach, only less than 1 hour is needed, 66 times faster than the centralized filter approach. The availability and reliability of network processing remain at the same high level.

• 전자공학회논문지B
• /
• 제30B권5호
• /
• pp.120-128
• /
• 1993

In this paper, an implementation of high-speed parallel processing system for neural network design on the multicomputer network is presented. Linear speedup expandability is increased by reducing the synchronization penalty and the communication overhead. Also, we presented the parallel processing models and their performance evaluation models for each of the parallization methods of the neural network. The results of the experiments for the character recognition of the neural network bases on the proposed system show that the proposed approach has the higher linear speedup expandability than the other systems. The proposed parallel processing models and the performance evaluation models could be used effectively for the design and the performance estimation of the neural network on the multicomputer network.