• Journal of Electrical Engineering and Technology
• /
• 제10권2호
• /
• pp.699-707
• /
• 2015

Although digit character recognition has got a significant improvement in recent years, it is still challenging to achieve satisfied result if the data contains an amount of distracting factors. This paper proposes a novel digit character recognition approach using a multi-layer hierarchical model, Hybrid Restricted Boltzmann Machines (HRBMs), which allows the learning architecture to be robust to background distracting factors. The insight behind the proposed model is that useful high-level features appear more frequently than distracting factors during learning, thus the high-level features can be decompose into hybrid hierarchical structures by using only small label information. In order to extract robust and compact features, a stochastic 0-1 layer is employed, which enables the model's hidden nodes to independently capture the useful character features during training. Experiments on the variations of Mixed National Institute of Standards and Technology (MNIST) dataset show that improvements of the multi-layer hierarchical model can be achieved by the proposed method. Finally, the paper shows the proposed technique which is used in a real-world application, where it is able to identify digit characters under various complex background images.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제11권2호
• /
• pp.785-804
• /
• 2017

Due to various factors such as postures, facial expressions and illuminations, face recognition by videos often suffer from poor recognition accuracy and generalization ability, since the within-class scatter might even be higher than the between-class one. Herein we address this problem by proposing a hierarchical cascaded classifier for video face recognition, which is a multi-layer algorithm and accounts for the misclassified samples plus their similar samples. Specifically, it can be decomposed into single classifier construction and multi-layer classifier design stages. In single classifier construction stage, classifier is created by clustering and the number of classes is computed by analyzing distance tree. In multi-layer classifier design stage, the next layer is created for the misclassified samples and similar ones, then cascaded to a hierarchical classifier. The experiments on the database collected by ourselves show that the recognition accuracy of the proposed classifier outperforms the compared recognition algorithms, such as neural network and sparse representation.

• 전자공학회논문지B
• /
• 제33B권8호
• /
• pp.50-59
• /
• 1996

In this paper, a new hierarchical motion estiamtion scheme using the wavelet transformed multi-resolution image layers is proposed. Compared with the full search motion estimation method, the existing hierarchical methods remarkably reduce the amount of the computation but their efficiencies are depreciated by the local minima problem. In order to solve the local minima problem, the multi-resolution image layers are composed using the wavelet transform and the number of layers participated in the motion estimation for a block is determined by considering of its low band energy and higher band energy on the first wavelet transformed layer. The ratio between higher band energy and low band energy of each block is evaluated and in the case of the blocks which include relatively large higher band energy, the motion estimation is carried out in the high resolution layer. Otherwise, all layers are used. The final motion vectors are obtained in the first wavelet transformed layer. So less bits for motion vectors are transmitted, and the decomposition of received image using inverse wavelet transform decreases the blocking effect.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제9권9호
• /
• pp.3468-3495
• /
• 2015

In this paper, we present a multi-chain based hierarchical topology control algorithm (MCHTC) for wireless sensor networks. In this algorithm, the topology control process using static clustering is divided into sensing layer that is composed by sensor nodes and multi-hop data forwarding layer that is composed by leader nodes. The communication cost and residual energy of nodes are considered to organize nodes into a chain in each cluster, and leader nodes form a tree topology. Leader nodes are elected based on the residual energy and distance between themselves and the base station. Analysis and simulation results show that MCHTC outperforms LEACH, PEGASIS and IEEPB in terms of network lifetime, energy consumption and network energy balance.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제11권12호
• /
• pp.6017-6037
• /
• 2017

In a multi-resolution image encoding system, the image is encoded into a single file as a layer of bit streams, and then it is transmitted layer by layer progressively to reduce the transmission time across a low bandwidth connection. This encoding scheme is also suitable for multiple decoders, each with different capabilities ranging from a handheld device to a PC. In our previous work, we proposed an edge adaptive hierarchical interpolation algorithm for multi-resolution image coding system. In this paper, we enhanced its compression efficiency by adding three major components. First, its prediction accuracy is improved using context adaptive error modeling as a feedback. Second, the conditional probability of prediction errors is sharpened by removing the sign redundancy among local prediction errors by applying sign flipping. Third, the conditional probability is sharpened further by reducing the number of distinct error symbols using error remapping function. Experimental results on benchmark data sets reveal that the enhanced algorithm achieves a better compression bit rate than our previous algorithm and other algorithms. It is shown that compression bit rate is much better for images that are rich in directional edges and textures. The enhanced algorithm also shows better rate-distortion performance and visual quality at the intermediate stages of progressive image transmission.

• 한국지능시스템학회논문지
• /
• 제23권3호
• /
• pp.226-231
• /
• 2013

의약품을 자동 포장하는 시스템에서는 캐니스터(Canister)에 해당 약을 정확히 보충할 수 있는 해당 약통과 캐니스터와의 일치 여부를 판단하는 정합 알고리즘이 필수적이다. 본 논문에서는 약화사고 방지를 위해 많은 종류의 약통을 분류하기 위한 분류 성능뿐만 아니라 실시간으로 처리할 수 있는 상 하 계층으로 구성된 계층적 신경회로망을 제안한다. 먼저 약통 정보를 나타내는 라벨 영상으로부터 다수의 저 차원 특징 벡터를 추출한다. 추출된 특징 벡터를 사용하여 하위계층의 다층 퍼셉트론(MLP, Multi-layer Perceptron) 신경회로망을 학습한다. 다음으로 학습된 MLP의 중간층 출력을 입력으로 사용하여 상위계층의 MLP를 학습한다. 100개의 약통에 대해 좌우 30도까지 회전한 영상에 대해 제안한 계층적 신경회로망의 분류 성능 시험과 실시간 연산처리 성능의 우수함을 보였다.

• 한국정밀공학회지
• /
• 제9권4호
• /
• pp.36-43
• /
• 1992

This paper presents a hierarchical real-time operation control and monitoring scheme of the FMS/CIM center which has been implemented at the Automation and Systems Research Institute of Seoul National University. The hierarchical structure of the whole scheme consists of three lavers. The upper layer is in charge of on-line scheduling, computer network control, shop-floor monitoring and command generation for AGV dispatching, machining, assembly, inspection, set-up, etc. The middle layer has six modules, which are installed in the FMS host computer with the upper layer and run on the multi-tasking basis. Each module is connected to one of six cell controllers distributed in the FMS model plant and transfers operation command down to each cell controller through the Ethernet/TCP-IP local area network. The lower layer is comprised of six cell control software modules for machining cell, assembly cell, inspection cell, set-up stations. AS/RS and AGV. Each cell controller reports the status of the manufacturing facilites to the middle layer as well as ecxecuting the appropriate sequence control of the manufacturing processes.

• 정보보호학회논문지
• /
• 제12권6호
• /
• pp.47-58
• /
• 2002

안전한 멀티캐스트 서비스와 관련하여 아키텍쳐, 키분배, 송신자 인증 등에 대한 연구가 활발히 이루어지고 있지만 서비스 거부 공격이나 권한 없는 멤버에 대한 멀티캐스트 서비스 접근을 통제할 수 있는 접근통제 기술에 대한 연구는 매우 미진한 상태이다. 멀티캐스트에서의 다단계 접근통제는 비밀 원격회의나 다양한 등급을 갖는 고객에 대한 차별된 멀티미디어 서비스를 제공하는데 응용할 수 있다. 실제 멀티캐스트 네트워크를 구성할 경우, 각각의 보안등급에 따라 서로 다른 가상 네트워크를 구성하게 된다. 그러나 기존 방식은 다중 접속 네트워크 환경에서의 불법접근을 효율적으로 막지 못하고 있고 다단계 접근통제 메커니즘을 제공하고 있지 않다. 따라서, 본 논문에서는 기존 멀티캐스트 접근통제 방식을 확장하여 네트워크 계층에서의 효율적인 계층형 접근통제 방식을 제안한다. 또한 어플리케이션 계층에서의 계층형 접근통제를 위한 계층키(hierarchical key) 분배 방식과 가입 및 탈퇴가 자유로운 동적 멀티캐스트 환경에서의 효율적인 계층키 갱신 방식을 제안한다.

• 한국멀티미디어학회논문지
• /
• 제18권9호
• /
• pp.1091-1097
• /
• 2015

Internet-of-Things (IoT) becomes one of the most promising future paradigms, which foresees enormous amounts of interoperable things and heterogeneous services. The goal of IoT is to enable all things connected and brings all kinds information and services to people. However, such a great deal of information may lead to cognitive overload or restrain in productivity of people. Thus, it is a necessity to build intelligent mechanisms to assist people in accessing the information or services they needed in a proactive manner. Most of previous related mechanisms are built on well-defined web services and lack of consideration of constrained resources. This paper suggests a services composition method by adapting a hierarchical model, which is a graph-based model composed of four layers: Context Layer, Event Layer, Service Layer and Device Layer. With a such multi-layer graph, service composition can be achieved by the iteration of layer by layer. Then, to evaluate the effectiveness of this proposed hierarchical model, a real-life emergency response dataset is applied and the experimental results are composed with the general probabilistic method and indicate that the proposed method is help for compositing multiple services while considering given context and constrained resources.

• 한국멀티미디어학회논문지
• /
• 제15권9호
• /
• pp.1093-1101
• /
• 2012

다시점 비디오는 3차원 정보를 표현하기 위한 영상으로 하나의 3차원 장면을 여러 시점에서 다수의 카메라로 촬영한 동영상이다. 영상들 사이에 존재하는 시간적 상관성과 화면간 상관성을 이용하는 다시점 비디오 부호화는 카메라의 수에 비례하여 데이터의 양이 늘어나기 때문에 계산량을 줄일 수 있는 다시점 비디오 부호화 기술이 필요하다. 본 논문에서는 다시점 비디오의 부호화 성능을 향상시키기 위한 효율적인 예측구조를 제안한다. 제안한 예측 구조는 다시점 비디오의 부호화 효율을 높이기 위하여 부호화되는 현재 화면과 현재 화면이 참조하는 참조 화면들과의 평균 거리, B계층 최대 인덱스 그리고 각 Bi 계층의 화면 수를 고려하였다. 제안한 예측 구조의 성능을 참조 예측 구조의 성능과 비교하였을 때 영상 화질 면에 있어서 제안한 예측 구조가 Fraunhofer-HHI의 계층적 B화면 구조보다 약 0.07~0.13 (dB) 성능 향상을 보였다. 발생되는 평균 초당 비트량에 있어서 제안한 예측 구조가 Fraunhofer-HHI의 계층적 B화면 구조보다 최대 6.5(Kbps) 감소하였다.