• 한국지능시스템학회논문지
• /
• 제16권1호
• /
• pp.1-7
• /
• 2006

본 논문에서는 6개의 특징점을 이용하는 가버 웨이블릿 신경망 기반 적응 표정인식 시스템을 제안한다. 특징 추출부를 포함하는 초기 네트워크의 구성은 Levenberg-Marquardt 기반의 학습방법이 사용되며, 따라서 특징 추출부 결정에 있어서 경험적 요소를 배재시킬 수 있다. 또한 새로운 사용자에 대한 적응 네트워크를 구성하기 위해서 개선된 보상함수를 가지는 Q-학습과, 비지도 퍼지 신경망 모델을 사용하였다. Q-학습을 통해서는 개인 사용자에 대해 분리도가 좋은 특징벡터를 얻을 수 있는 가버필터 세트를 얻을 수 있으며, 퍼지 신경망을 통해서는 사용자의 얼굴변화에 맞게 인식기를 변화시킬 수 있다. 따라서 제안된 시스템은 사용자의 얼굴변화를 따라갈 수 있는 좋은 적응 성능을 보이고 있다.

• 한국산학기술학회논문지
• /
• 제6권2호
• /
• pp.149-158
• /
• 2005

본 논문은 이온 선택성 전극을 모듈화한 MACS를 사용하여 시스템의 크기를 축소할 수 있었고, PDA를 사용함으로써 측정된 데이터를 장소에 구애받지 않고 분석할 수 있는 휴대용 전자혀 시스템을 개발하였다. MACS는 ${NH_4}^+$, $Na^+$, $Cl^-$, ${NO_3}^-$, $K^+$, $Ca^{2+}$, $Na^+$, pH의 7종의 이온 선택성 전극을 이용하여 구성하였으며, 초기화 및 교정과정과 완충용액에 의한 안정화 과정을 거친 후 MACS로 시료에 대한 각각의 이온선택성 전극의 변화를 측정한다. 이렇게 각 전극으로부터 측정된 데이터를 이용하여 신경회로망 알고리즘으로 측정된 시료의 종류를 구분할 수 있다. 실험은 분류가 어렵다고 알려진 고급양주와 저급양주를 분류하는 것으로 진행되었으며, 성공적이며 우수한 실험 결과를 얻었다 이로부터 사용된 알고리즘이 휴대용 전자혀 시스템에 적절히 사용될 수 있음을 밝혔으며, 실제 휴대용 전자혀 시스템에 간단한 학습에 의해 적용될 수 있을 것으로 생각된다.

• Coupled systems mechanics
• /
• 제7권2호
• /
• pp.197-209
• /
• 2018

Fiber-reinforced concrete (FRC) is a material with increasing application in civil engineering. Here it is assumed that the material consists of a great number of rather small fibers embedded into the concrete matrix. It would be advantageous to predict the mechanical properties of FRC using nondestructive testing; unfortunately, many testing methods for concrete are not applicable to FRC. In addition, design methods for FRC are either inaccurate or complicated. In three-point bending tests of FRC prisms, it has been observed that fiber reinforcement does not break but simply pulls out during specimen failure. Following that observation, this work is based on an assumption that the main components of a simple and rather accurate FRC model are mechanical properties of the concrete matrix and fiber pullout force. Properties of the concrete matrix could be determined from measurements on samples taken during concrete production, and fiber pullout force could be measured on samples with individual fibers embedded into concrete. However, there is no clear relationship between measurements on individual samples of concrete matrix with a single fiber and properties of the produced FRC. This work presents an inverse model for FRC that establishes a relation between parameters measured on individual material samples and properties of a structure made of the composite material. However, a deterministic relationship is clearly not possible since only a single beam specimen of 60 cm could easily contain over 100000 fibers. Our inverse model assumes that the probability density function of individual fiber properties is known, and that the global sample load-displacement curve is obtained from the experiment. Thus, each fiber is stochastically characterized and accordingly parameterized. A relationship between fiber parameters and global load-displacement response, the so-called forward model, is established. From the forward model, based on Levenberg-Marquardt procedure, the inverse model is formulated and successfully applied.

• 한국정보통신학회:학술대회논문집
• /
• 한국해양정보통신학회 2008년도 춘계종합학술대회 A
• /
• pp.229-232
• /
• 2008

A Neural network classification of human activity data is presented. The data acquisition system involves a tri-axial accelerometer in wireless sensor network environment. The wireless ad-hoc system has the advantage of small size, convenience for wearability and cost effectiveness. The system can further improve the range of user mobility with the inclusion of ad-hoc environment. The classification is based on the frequencies of the involved activities. The most significant Fast Fourier coefficients, of the acceleration of the body movement, are used for classification of the daily activities like, Rest walk and Run. A supervised learning approach is used. The work presents classification accuracy with the available fast batch training algorithms i.e. Levenberg-Marquardt and Resilient back propagation scheme is used for training and calculation of accuracy.

• 상하수도학회지
• /
• 제19권4호
• /
• pp.395-403
• /
• 2005

It is very important to forecast water supply for reasonal operation and management of water utilities. In this paper, water supply forecasting models using artificial intelligence are developed. Artificial intelligence models shows better results by using Temperature(t), water supply discharge (t-1) and water supply discharge (t-2), which are expressed by neural network(LMNNWS; Levenberg-Marquardt Neural Network for Water Supply, MDNNWS; MoDular Neural Network for Water Supply) and neuro fuzzy(ANASWS; Adaptive Neuro-Fuzzy Inference Systems for Water Supply). ANFISWS model which is applied for water supply forecasting shows stable application to the variable water supply data. As results, MDNNWS model shows the highest overall accuracy among proposed water supply forecasting models and the lowest estimation error with the order of ANFISWS, LMNNWS model.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2003년도 ICCAS
• /
• pp.762-766
• /
• 2003

In this paper, we present a new approach to solve the problem of estimating the camera 3-D location and orientation from a matched set of 3-D model and 2-D image features. An iterative least-square method is used to solve both rotation and translation simultaneously. Because conventional methods that solved for rotation first and then translation do not provide good solutions, we derive an error equation using roll-pitch-yaw angle to present the rotation matrix. To minimize the error equation, Levenberg-Marquardt algorithm is introduced with uniform sampling strategy of rotation space to avoid stuck in local minimum. Experimental results using real images are presented.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2004년도 학술대회 논문집 정보 및 제어부문
• /
• pp.39-42
• /
• 2004

We have designed a portable electronic nose(e-nose) system using an array of commercial gas sensors for recognition and analyzing the various odours. In this paper, we have implemented a portable e-nose system using an array gas sensors and personal digital assistants(PDA) for recognizing and analyzing volatile organic compounds(VOCs) in the field. Field screening for pollutants has been a target of instrumental development for number of year. A portable e-nose system can be substantial benefit to rapidly localize the spacial extent of a pollution or to find pollutants source. And, by using PDA, E-nose have a better function such as the easy user-interface and data transfer by internet from on- site to remote computer. We adapted the Levenberg-Marquardt algorithm based on the back-propagation and proposed the method that could be predicted concentration levels of VOCs gases after classification by separating neural network into two parts.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2001년도 제14회 신호처리 합동 학술대회 논문집
• /
• pp.415-418
• /
• 2001

The automatic construction of large, high-resolution image mosaics is an active area of reasearch in the fields of photogrammetry, computer vision, image processing, and computer graphics. In this study, we describe a automatic mosaicing method that makes a panorama from images by placing camera in a emitted-grid. In the images captured by cameras, there must be a matched area and the area is in the particular area of the image. Initial transformation matrix, there(ore, is calculated from points searched in the partial area. It is possible to find best transformation matrix by Levenberg-Marquardt method. Finally, each images are multiplied by blending function and stitched by the transformation matrix to complete panoramic image.

• Journal of Advanced Marine Engineering and Technology
• /
• 제27권5호
• /
• pp.641-647
• /
• 2003

The use of infrared range-finder sensor as the environment recognition system for mobile robot have the advantage of low sensing cost compared with the use of other vision sensor such as laser finder CCD camera. However, it is not easy to find the previous works on the use of infrared range-finder sensor for a mobile robot because of the non-linear characteristic of that. This paper describes the error due to non-linearity of a sensor and the correction of it using neural network. The neural network consists of multi-layer perception and Levenberg-Marquardt algorithm is applied to learning it. The effectiveness of the proposed algorithm is verified from experiment.

• Applied Microscopy
• /
• 제43권1호
• /
• pp.1-8
• /
• 2013

We developed an improved method for tilt series alignment with fiducial markers in electron tomography. Based on previous works regarding alignment, we adapted the Levenberg-Marquardt method to solve the nonlinear least squares problem by incorporating a new formula for the alignment model. We also suggested a new method to estimate the initial value for inversion with higher accuracy. The proposed approach was applied to geopolymers. A better alignment of the tilt series was achieved than that by IMOD S/W. The initial value estimation provided both stability and a good rate of convergence since the new method uses all marker positions, including those partly covering the tilt images.