• Journal of Korea Multimedia Society
• /
• v.14 no.12
• /
• pp.1544-1548
• /
• 2011

A process of choosing a subset of original features, so called feature selection, is considered as a crucial preprocessing step to image processing applications. There are already large pools of techniques developed for machine learning and data mining fields. In this paper, basically two methods, non-feature selection and feature selection, are investigated to compare their predictive effectiveness of classification. Color co-occurrence feature is used for defining image features. Standard Sequential Forward Selection algorithm are used for feature selection to identify relevant features and redundancy among relevant features. Four color spaces, RGB, YCbCr, HSV, and Gaussian space are considered for computing color co-occurrence features. Gray-level image feature is also considered for the performance comparison reasons. The experimental results are presented.

• Proceedings of the KIEE Conference
• /
• 1999.11c
• /
• pp.824-826
• /
• 1999

In this paper, we design an Fuzzy-Neural Networks(FNN) by means of divisions of fuzzy input space with multi-input variables. Fuzzy input space of Yamakawa's FNN is divided by each separated input variable, but that of the proposed FNN is divided by mutually combined input variables. The membership functions of the proposed FNN use both triangular and gaussian membership types. The parameters such as apexes of membership functions, learning rates, momentum coefficients, weighting value, and slope are adjusted using genetic algorithms. Also, an aggregate objective function(performance index) with weighting value is utilized to achieve a sound balance between approximation and generalization abilities of the model. To evaluate the performance of the proposed model, we use the data of sewage treatment process.

• Journal of the Korean Society of Manufacturing Technology Engineers
• /
• v.23 no.2
• /
• pp.157-163
• /
• 2014

In this paper, a robust MRAC (model reference adaptive control) scheme is applied to control an electrohydraulic positioning system under various loads. The inverse dead-zone compensator in the control system cancels out the dead-zone response, and an integrator added to the controller provides good position-tracking ability. LQG/LTR (linear quadratic Gaussian control with loop transfer recovery) closed-loop model is used as the reference model for learning the MRAC system. LQG/LTR provides a systematic technique to design the linear controller that optimizes the objective function using some compromise between the control effort and the system performance in the frequency domain. Different external load tests are performed to investigate the effectiveness of the designed MRAC system in real time. The experimental results show that the tracking performance of the proposed system is highly accurate, which offers considerable robustness even with a large change in the load.

• Proceedings of the IEEK Conference
• /
• 1999.11a
• /
• pp.368-371
• /
• 1999

In this paper, we propose a neural network processor architecture with on-chip learning and with reconfigurability according to the data dependencies of the algorithm applied. For the neural network model applied, the proposed architecture can be configured into either SIMD or SRA(Systolic Ring Array) without my changing of on-chip configuration so as to obtain a high throughput. However, changing of system configuration can be controlled by user program. To process activation function, which needs amount of cycles to get its value, we design it by using PWL(Piece-Wise Linear) function approximation method. This unit has only single latency and the processing ability of non-linear function such as sigmoid gaussian function etc. And we verified the processing mechanism with EBP(Error Back-Propagation) model.

• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 2003.04a
• /
• pp.357-365
• /
• 2003

The major problems with the conventional neural network, especially Back Propagation Neural Network, arise from the necessity of many training data for neural network learning and ambiguity in the relation of neural network structure to the convergence of solution. In this paper, the PNN is used as a pattern classifier to detect the damage of structure to avoid those drawbacks of the conventional neural network. In the PNN-based pattern classification problems, the probability density function for patterns is usually assumed by Gaussian distribution. But, in this paper, several probability density functions are investigated in order to select the most approriate one for structural damage assessment.

• Journal of Korea Multimedia Society
• /
• v.20 no.2
• /
• pp.108-114
• /
• 2017

Lip reading is a field of image processing to assist the process of sound recognition. In some environment, the capture of sound signal usually has significant noise and therefore, the recognition rate of sound signal decreases. Lip reading can be a good feature for the increase of recognition rates. Conventional lip extraction methods have been proposed widely. Maia et. al. proposed a method by the sum of Cr and Cb. However, there are two problems as follows: the point with maximum saturation is not always regarded as lips region and the inner part of lips such as oral cavity and teeth can be classified as lips. To solve these problems, this paper proposes a method which adopts the histogram-based classifier for the extraction of lips region. The proposed method consists of two stages, learning and test. The amount of computation is minimized because this method has no color conversion. The performance of proposed method gives 66.8% of detection rate compared to 28% of conventional ones.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.53 no.7
• /
• pp.536-543
• /
• 2004

In this paper, we proposed a new method of modeling a neuro-fuzzy system using a hybrid clustering algorithm. The initial parameters and the number of clusters of the proposed system are optimally chosen simultaneously with respect to the process of regression, which is a unique characteristics of the proposed system. The proposed algorithm presented in this work improves the overall performance of the proposed a neuro-fuzzy system by choosing a proper number of clusters adaptively according the characteristics of given data. The process of clustering is performed by deciding on the number of classes, which yields the property of convergence of the system. In experiments, the superiority of the proposed neuro-fuzzy system is demonstrated, especially the process of optimizing parameters and clustering of learning speed.

• International Journal of Control, Automation, and Systems
• /
• v.6 no.1
• /
• pp.109-118
• /
• 2008

We present two estimators for discrete non-Gaussian and nonstationary probability density estimation based on a dynamic Bayesian network (DBN). The first estimator is for off line computation and consists of a DBN whose transition distribution is represented in terms of kernel functions. The estimator parameters are the weights and shifts of the kernel functions. The parameters are determined through a recursive learning algorithm using maximum likelihood (ML) estimation. The second estimator is a DBN whose parameters form the transition probabilities. We use an asymptotically convergent, recursive, on-line algorithm to update the parameters using observation data. The DBN calculates the state probabilities using the estimated parameters. We provide examples that demonstrate the usefulness and simplicity of the two proposed estimators.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1998.10a
• /
• pp.474-478
• /
• 1998

A new approach for the decision feedback equalizer(DFE) based on the back-propagation neural networks is described. We propose the method of optimal structure for back-propagation neural networks model. In order to construct an the optimal structure, we first prescribe the bounds of learning procedure, and the, we employ the method of incrementing the number of input neuron by utilizing the derivative of the error with respect to an hidden neuron weights. The structure is applied to the problem of adaptive equalization in the presence of inter symbol interference(ISI), additive white Gaussian noise. From the simulation results, it is observed that the performance of the propose neural networks based decision feedback equalizer outperforms the other two in terms of bit-error rate(BER) and attainable MSE level over a signal ratio and channel nonlinearities.

• Journal of KIEE
• /
• v.11 no.1
• /
• pp.1-7
• /
• 2001

Various methods for blind source separation (BSS) are based on independent component analysis (ICA) which can be viewed as a nonlinear extension of principal component analysis (PCA). Most existing ICA methods require certain nonlinear functions (which leads to higher-order statistics) depending on the probability distributions of sources, whereas PCA is a linear learning method based on second-order statistics. In this paper we show that the PCA can be applied to the task of BBS, provided that source are spatially uncorrelated but temporally correlated. Since the resulting method is based on only second-order statistics, it avoids the nonlinear function and is able to separate mixtures of several colored Gaussian sources, in contrast to the conventional ICA methods.