• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.32 no.4
• /
• pp.24-33
• /
• 2004

In this paper, the helicopter flight control system using online adaptive neural networks which have the universal function approximation property is considered. It is not compensation for modeling errors but approximation two functions required for feedback linearization control action from input/output of the system. To guarantee the tracking performance and the stability of the closed loop system replaced two nonlinear functions by two neural networks, weight update laws are provided by Lyapunov function and the simulation results in low speed flight mode verified the performance of the control system with the neural networks.

• Journal of Digital Contents Society
• /
• v.18 no.3
• /
• pp.559-565
• /
• 2017

A lot of studies have been conducted to recognize the motion state or behavior of the user using the acceleration sensor built in the smartphone. In this paper, we applied the neural networks to the 3-axis acceleration information of smartphone to study human behavior. There are performance issues in applying time series data to neural networks. We proposed a multi-temporal neural networks which have trained three neural networks with different time windows for feature extraction and uses the output of these neural networks as input to the new neural network. The proposed method showed better performance than other methods like SVM, AdaBoot and IBk classifier for real acceleration data.

• 제어로봇시스템학회:학술대회논문집
• /
• 1997.10a
• /
• pp.215-218
• /
• 1997

A dynamic back-propagation neural network is addressed for adaptive neural control system to approximate non-linear control system rather than static networks. It has the capability to represent the approximation of nonlinear system without mathematical analysis and to carry out the on-line learning algorithm for real time application. The simulated results show fast tracking capability and adaptive response by using dynamic back-propagation neurons.

• Proceedings of the Korean Information Science Society Conference
• /
• 1998.10c
• /
• pp.357-359
• /
• 1998

무선 통신 환경에서 기지국 안테나를 떠난 전파가 수신안테나에 도달하는 과정 중에 발생하는 전파 손실은 매우 복잡한 비선형 함수이다. 본 논문에서는 신경회로망을 사용한 전파 손실 모델을 제안하고, 3차원 지형 데이터를 이용하여 전파 환경을 반영할 수 있는 특징을 추출하여 이를 신경회로망에 적용함으로써 전파손실 예측모델을 생성하는 방법을 소개한다. 각 필드 측정 데이터에 대한 특징 값을 이용하여 신경회로망을 학습하여 예측모델을 완성한다. 또한, 서울 도심 지역의 실제 PCS 서비스 환경에 대한 실험결과를 통해 제안하는 모델의 우수성을 보인다.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1998.10a
• /
• pp.263-266
• /
• 1998

이 논문에서는 비선형 Channel의 등화기를 설계하기 위해 새로운 구조를 갖는 신경회로망을 제안하였다. 비선형 Channel의 동적 특성을 제대로 학습하기 위해 새로운 신경회로망은 은닉층 노드의 출력이 은닉층의 입력으로 되먹임되는 구조를 갖는다. 또한 이 논문에서는 제안한 신경회로망의 구조에 알맞는 학습 알고리즘을 제안하였다. 제안한 신경회로망과 학습 알고리즘의 성능은 Computer simulation을 통해 보였고, 그 결과는 기존의 Channel 등화기를 사용했을 경우보다 나은 결과를 보여 주었다.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2011.10a
• /
• pp.613-615
• /
• 2011

본 논문에서는 켑스트럼 계수를 이용하여 음성인식을 하는 알고리즘을 제안한다. 본 논문에서 제안하는 방법은 사람이 발성한 음성을 두 영역의 켑스트럼 계수로 분리한 후에, 신경회로망을 사용하여 음성인식을 하는 방법이다. 본 논문에서 제안하는 신경회로망은 오차가 거의 없어지는 일정 기간 동안 네트워크를 학습시킨 후에 신경회로망의 학습 데이터와는 다른 새로운 음성이 신경회로망에 입력된 경우에 대하여 각 음성 구간에서 분류가 가능한 모음검출을 위한 음성인식 시스템을 제안한다.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2006.11a
• /
• pp.197-202
• /
• 2006

본 논문에서는 퍼지 신경회로망을 사용하여 영상에서 물체를 배경으로부터 추출해내는 방법을 제시하였다. 퍼지 신경회로망의 vigilance parameter를 조정하여 영상을 2개의 클래스로 분류하고, 물체 영역과 배경영역의 Cb와 Cr의 대표값을 추출하였다. 제안한 방법을 사용하여 물체색상의 위치 및 크기와 밝기에 상관없이 물체영역을 추출하였다.

• Journal of the Korean Institute of Telematics and Electronics C
• /
• v.36C no.4
• /
• pp.78-86
• /
• 1999

In this paper, a distributed arithmetic digital neural network with learning and testing phase implemented in a body has been studied. The proposed technique is based on the two facts; one is that the weighting coefficients adjusted will be stored in registers without shift, because input values or input patterns are not changed while learning and the other is that the input patterns stored in registers are not changed while testing. The proposed digital neural network is simulated by hardware description language such as VHDL and verified the performance that the neural network was applied to the recognition of seven-segment. To verify proposed neural networks, we compared the learning process of modified perceptron learning algorithm simulated by software with VHDL for 7-segment number recognizer. The results are as follows: There was a little difference in learning time and iteration numbers according to the input pattern, but generally the iteration numbers are 1000 to 10000 and the learning time is 4 to 200$\mu\textrm{s}$. So we knew that the operation of the neural network is learned in the same way with the learning of software simulation, and the proposed neural networks are properly operated. And also the implemented neural network can be built with less amounts of components compared with board system neural network.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.41 no.3
• /
• pp.55-65
• /
• 2004

When a neural network is used to solve a given problem it is necessary to match the complexity of the network to that of the problem because the complexity of the network significantly affects its learning capability and generalization performance. Thus, it is desirable to have an algorithm that can find appropriate network structures in a self-organizing way. This paper proposes algorithms which automatically organize feed forward multi-layer neural networks with sigmoid hidden neurons for given problems. Using both constructive procedures and pruning procedures, the proposed algorithms try to find the near optimal network, which is compact and shows good generalization performance. The performances of the proposed algorithms are tested on four function regression problems. The results demonstrate that our algorithms successfully generate near-optimal networks in comparison with the previous method and the neural networks of fixed topology.

• Journal of KIISE:Software and Applications
• /
• v.30 no.3_4
• /
• pp.326-338
• /
• 2003

The objective of a neural network design and model selection is to construct an optimal network with a good generalization performance. However, training data include noises, and the number of training data is not sufficient, which results in the difference between the true probability distribution and the empirical one. The difference makes the teaming parameters to over-fit only to training data and to deviate from the true distribution of data, which is called the overfitting phenomenon. The overfilled neural network shows good approximations for the training data, but gives bad predictions to untrained new data. As the complexity of the neural network increases, this overfitting phenomenon also becomes more severe. In this paper, by taking statistical viewpoint, we proposed an integrative process for neural network design and model selection method in order to improve generalization performance. At first, by using the natural gradient learning with adaptive regularization, we try to obtain optimal parameters that are not overfilled to training data with fast convergence. By adopting the natural pruning to the obtained optimal parameters, we generate several candidates of network model with different sizes. Finally, we select an optimal model among candidate models based on the Bayesian Information Criteria. Through the computer simulation on benchmark problems, we confirm the generalization and structure optimization performance of the proposed integrative process of teaming and model selection.