• 한국음향학회지
• /
• 제26권6호
• /
• pp.259-268
• /
• 2007

본 논문에서는 HMM 기반 통계적인 대역폭 확장(Bandwidth Extension, BWE) 방법의 개선에 대해 다룬다. 이를 위해 우선, HMM 모델 학습을 위한 기존의 Jax의 학습법과 일반적인 Baum-Welch 학습법의 관계를 비교 검토하고, Jax의 학습법의 한계점 및 문제점을 검토한다. 그리고 이를 바탕으로 Baum-Welch학습법을 이용한 새로운 HMM 기반 BWE 방법을 제시한다. 결론적으로, Baum-Welch 학습법은 Jax의 학습법의 일반화된 형태로 볼 수 있으며, 보다 유연하고 적응적인 학습능력을 가진 알고리즘임을 알 수 있다. 따라서 학습 데이터에 대한 보다 정확한 HMM 모델링이 가능하며 아울러, 이와 같이 개선된 HMM 모델을 활용함으로써 BWE 시스템의 성능향상을 가져 올 수 있었다. 실험결과에 의하면, 제시된 새로운 방법이 기존의 Jax의 방법에 비해 실험의 모든 경우에서 우수한 성능을 보임을 알 수 있다. 주어진 실험조건하에서 근제곱평균(root-mean-square, RMS) 로그 스펙트럴 왜곡(Log Spectral Distortion, LSD) 값이 전체적으로 평균 0.52dB 그리고, 최소 0.31dB에서 최대 0.8dB까지 개선되었다.

• Water Engineering Research
• /
• 제1권2호
• /
• pp.129-136
• /
• 2000

Predicting water levels is a difficult task because a lot of uncertainties are included. Therefore the neural network which is appropriate to such a problem, is introduced. One day ahead forecasting of river stage in the Bocheong River is carried out by using the neural network model. Historical water levels at Snagye gauging point which is located at the downstream of the Bocheong River and average rainfall of the Bocheong River basin are selected as training data sets. With these data sets, the training process has been done by using back propagation algorithm. Then waters levels in 1997 and 1998 are predicted with the trained algorithm. To improve the accuracy, a filtering method is introduced as predicting scheme. It is shown that predicted results are in a good agreement with observed water levels and that a filtering method can overcome the lack of training patterns.

• The Journal of the Acoustical Society of Korea
• /
• 제23권4E호
• /
• pp.108-111
• /
• 2004

This paper has interest in the effect of a fine frequency offset, defined in ITU-T G.225, to the training performance of an adaptive equalizer. This paper uses Hilbert filter in configuring a transmission system model in order to let it get a frequency offset. Also additive white Gaussian noise and band-limited filter are considered. The signal received from the above transmission system applies to an adaptive equalizer with LMS algorithm, and its training procedures are investigated. As a result, we could find that even small fine frequency offset can severely deteriorate training performance of adaptive algorithm.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2017년도 춘계학술발표대회
• /
• pp.787-790
• /
• 2017

Data quality is the main issue in the classification problems; generally, the presence of noisy instances in the training dataset will not lead to robust classification performance. Such instances may cause the generated decision tree to suffer from over-fitting and its accuracy may decrease. Decision trees are useful, efficient, and commonly used for solving various real world classification problems in data mining. In this paper, we introduce a preprocessing technique to improve the classification accuracy rates of the C4.5 decision tree algorithm. In the proposed preprocessing method, we applied the naive Bayes classifier to remove the noisy instances from the training dataset. We applied our proposed method to a real e-commerce sales dataset to test the performance of the proposed algorithm against the existing C4.5 decision tree classifier. As the experimental results, the proposed method improved the classification accuracy by 8.5% and 14.32% using training dataset and 10-fold crossvalidation, respectively.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2005년도 ICCAS
• /
• pp.2521-2525
• /
• 2005

In this paper, we presented a predictive control technique, which is based on wavelet neural network (WNN), for the control of chaotic systems whose precise mathematical models are not available. The WNN is motivated by both the multilayer feedforward neural network definition and wavelet decomposition. The wavelet theory improves the convergence of neural network. In order to design predictive controller effectively, the WNN is used as the predictor whose parameters are tuned by error between the output of actual plant and the output of WNN. Also the training method for the finding a good WNN model is the Extended Kalman algorithm which updates network parameters to converge to the reference signal during a few iterations. The benefit of EKF training method is that the WNN model can have better accuracy for the unknown plant. Finally, through computer simulations, we confirmed the performance of the proposed control method.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 1992년도 한국자동제어학술회의논문집(국내학술편); KOEX, Seoul; 19-21 Oct. 1992
• /
• pp.619-623
• /
• 1992

In pattern recognition, the conventional neural networks contain a large number of weights and require considerable training times and preprocessor to classify a transformed patterns. In this paper, we propose a constrained pattern recognition method which is insensitive to rotation of input pattern by various degrees and does not need any preprocessing. Because these neural networks can not be trained by the conventional training algorithm such as error back propagation, a novel training algorithm is suggested. As such a system is useful in problem related to calssify overse side and reverse side of 500 won coin. As an illustrative example, identification problem of overse and reverse side of 500 won coin is shown.

• 한국기계가공학회지
• /
• 제16권6호
• /
• pp.125-132
• /
• 2017

Generally, in machining operations, the required machining performance can be obtained by properly combining several machining parameters properly. In this research, we construct a simulation model, which that predicts the relationship between the input variables and output variables in the turning operation. Input variables necessary for the turning operation include cutting speed, feed, and depth of cut. Surface roughness and electrical current consumption are used as the output variables. To construct the simulation model, an Artificial Neural Network (ANN) is employed. With theIn ANN, training is necessary to find appropriate weights, and the Ant Colony Optimization (ACO) technique is used as a training tool. EspeciallyIn particular, for the continuous domain, ACOR is adopted and athe related algorithm is developed. Finally, the effects of the algorithm on the results are identified and analyzsed.

• 제어로봇시스템학회논문지
• /
• 제3권6호
• /
• pp.594-601
• /
• 1997

A new algorithm is proposed for training the multilayer perceptrion(MLP) in pattern classification problems to accelerate the learning speed. It is shown that the sigmoid activation function of the output node can have deterimental effect on the performance of learning. To overcome this detrimental effect and to use the information fully in supervised learning, an objective function for binary modes is proposed. This objective function is composed with two new output activation functions which are selectively used depending on desired values of training patterns. The effect of the objective function is analyzed and a training algorithm is proposed based on this. Its performance is tested in several examples. Simulation results show that the performance of the proposed method is better than that of the conventional error back propagation (EBP) method.

• 한국지능시스템학회:학술대회논문집
• /
• 한국퍼지및지능시스템학회 2001년도 춘계학술대회 학술발표 논문집
• /
• pp.105-108
• /
• 2001

In this paper, the learning algorithm called advanced expanded and truncate algorithm(AETLA) is proposed to training multilayer binary neural network to approximate binary to binary mapping. AETLA used merit of ETL and MTGA learning algorithm. We proposed to new learning algorithm to decrease number of hidden layer. Therefore, learning speed of the proposed AETLA learning algorithm is much faster than other learning algorithm.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제7권11호
• /
• pp.2720-2736
• /
• 2013

Multi-view face detection has become an active area for research in the last few years. In this paper, a novel multi-view human face detection algorithm based on improved real Adaboost is presented. Real Adaboost algorithm is improved by weighted combination of weak classifiers and the approximately best combination coefficients are obtained. After that, we proved that the function of sample weight adjusting method and weak classifier training method is to guarantee the independence of weak classifiers. A coarse-to-fine hierarchical face detector combining the high efficiency of Haar feature with pose estimation phase based on our real Adaboost algorithm is proposed. This algorithm reduces training time cost greatly compared with classical real Adaboost algorithm. In addition, it speeds up strong classifier converging and reduces the number of weak classifiers. For frontal face detection, the experiments on MIT+CMU frontal face test set result a 96.4% correct rate with 528 false alarms; for multi-view face in real time test set result a 94.7 % correct rate. The experimental results verified the effectiveness of the proposed approach.