• 한국HCI학회:학술대회논문집
• /
• 2009.02a
• /
• pp.418-421
• /
• 2009

This paper presents a human-machine interaction based on a realtime upper limb motion prediction method using surface electromyography (sEMG). The motions were predicted using an artificial neural network algorithm and sEMG signals which are acquired from five muscles, and then a manipulator was controlled to follow after the predicted motions. Upper limb motions were restricted to 2D vertical plane with the contact condition between a user and an end-effector of manipulator. In order to demonstrate the feasibility of the proposed method, experiments using developed method and using a goniometer were performed. The results showed that the proposed real-time motion prediction method can be implemented a human-machine interaction system.

• Journal of the Korean Institute of Illuminating and Electrical Installation Engineers
• /
• v.24 no.4
• /
• pp.33-42
• /
• 2010

The induction motor drive applied to field oriented control is widely used in industry applications. However, it is deceased performance and authenticity by saturation, temperature changing, disturbance and parameters changing because modeling of induction motor is nonlinear and complex. In order to control variable speed operation, conventional PI-like controllers are commonly used. These controllers provide limited good performance over a wide range of operation, even under ideal field oriented conditions. This paper proposes self tuning PI controller based on fuzzy-neural network(FNN)-PI controller that is implemented using fuzzy control, neural network, and adaptive fuzzy controller(AFC). Also, this paper proposes estimation of speed using ANN. The proposed control algorithm is applied to induction motor drive system using FNN-PI, AFC and ANN controller. Also, this paper proposes the anlysis results to verify the effectiveness of controller.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2018.05a
• /
• pp.466-467
• /
• 2018

Recently, artificial intelligence has been applied to various fields such as image recognition, image recognition speech recognition, and natural language processing, and interest in Deep Learning technology is increasing. Many researches on Convolutional Neural Network(CNN), which is one of the most representative algorithms among Deep Learning, have strong advantages in image recognition and classification and are widely used in various fields. In this paper, we propose a new network structure that transforms the general CNN structure. A typical CNN structure consists of a convolution layer, ReLU layer, and a pooling layer. Therefore in this paper, We intend to construct a new network by adding fully connected layer inside a general CNN structure. This modification is intended to increase the learning and accuracy of the convoluted image by including the generalization which is an advantage of the neural network.

• Journal of The Korean Association of Information Education
• /
• v.14 no.4
• /
• pp.527-536
• /
• 2010

Ubiquitous computing and mobile technologies provide much scope for designing innovative learning experiences that can take place in a variety of indoor and outdoor settings, such as botanical gardens, parks, museums and classrooms. In this paper, we present our own innovative work for bridging indoor and outdoor field-trip learning activities with the support of RFID with reader device on ubiquitous environments. The function of our system is based on strategically located RFID tags placed on objects around settings which are identified using the mobile device installed to an RFID reader. The mobile device reads the RFID tag and sends the learner's request to the field-trip learning system. And learners can be learning activities according to theirs learning levels. We also focus on the use of intelligent agents to customize learning contents for individual learners. The result of our experiment indicates that learner enjoyed learning where mobile devices are used in the system, supporting the learning activities in the context of which they are taking place.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.23 no.5
• /
• pp.473-478
• /
• 2013

In this paper, we develop the Heavy Rain Advisory Decision Model based on intelligent neuro-fuzzy algorithm RBFNNs by using KLAPS(Korea Local Analysis and Prediction System) Reanalysis data. the prediction ability of existing heavy rainfall forecasting systems is usually affected by the processing techniques of meteorological data. In this study, we introduce the heavy rain forecast method using the pre-processing techniques of meteorological data are in order to improve these drawbacks of conventional system. The pre-processing techniques of meteorological data are designed by using point conversion, cumulative precipitation generation, time series data processing and heavy rain warning extraction methods based on KLAPS data. Finally, the proposed system forecasts cumulative rainfall for six hours after future t(t=1,2,3) hours and offers information to determine heavy rain advisory. The essential parameters of the proposed model such as polynomial order, the number of rules, and fuzzification coefficient are optimized by means of Differential Evolution.

• The Journal of the Acoustical Society of Korea
• /
• v.19 no.1
• /
• pp.4-10
• /
• 2000

The work presented in this paper is about a postprocessor, which improves the performance of automatic speech segmentation system by correcting the phoneme boundary errors. We propose a postprocessor that reduces the range of errors in the auto labeled results that are ready to be used directly as synthesis unit. Starting from a baseline automatic segmentation system, our proposed postprocessor trains the features of hand labeled results using multi-layer perceptron(MLP) algorithm. Then, the auto labeled result combined with MLP postprocessor determines the new phoneme boundary. The details are as following. First, we select the feature sets of speech, based on the acoustic phonetic knowledge. And then we have adopted the MLP as pattern classifier because of its excellent nonlinear discrimination capability. Moreover, it is easy for MLP to reflect fully the various types of acoustic features appearing at the phoneme boundaries within a short time. At the last procedure, an appropriate feature set analyzed about each phonetic event is applied to our proposed postprocessor to compensate the phoneme boundary error. For phonetically rich sentences data, we have achieved 19.9 % improvement for the frame accuracy, comparing with the performance of plain automatic labeling system. Also, we could reduce the absolute error rate about 28.6%.

• Journal of the Korean Society for Precision Engineering
• /
• v.15 no.8
• /
• pp.17-25
• /
• 1998

Aluminum alloys have low density, relatively high strength and yield strength, good plasticity, good machinability, and high corrosion and acid resistance. Therefore, they are suitable for large containers for the food, chemical and other industries. Large containers are often bodies of revolution consisting of shell courses, stiffening rings, heads and other elements joined by annular welds. Larger containers have longer welds and require greater leak-tightness and higher weld mechanical properties. The LNG tank consists of aluminum plates with various sizes, so its construction should by divided by several sections. Moreover, each section has its own sub-section consisted of several aluminum plates. To guarantee the quality of huge LNG tank, therefore, the precise control of plate dimension should by urgently needed in conjunction with the appropriate selection of process parameters such as cutting speed, depth of cut, rotational speed and so on. In this paper, a manufacturing system was developed to implement automatic circular tracking in height direction and automatic circular interpolation in depth of cut direction. Also, the neural network based on the backpropagation algorithm was used to predict the cutting quality and motor load related with the life time of the developed system. It was revealed that the manufacturing system and the neural network could be effectively applied to the bevelling process and to predict the quality of machined area and the motor load.

• Transactions of the Korean Society of Mechanical Engineers
• /
• v.17 no.1
• /
• pp.101-111
• /
• 1993

Many processes such as machining, injection-moulding and metal-forming are usually operated by hydraulic servo-systems. The dynamic characteristics of these systems are complex and highly non-linear and are often subjected to the uncertain external disturbances associated with the processes. Consequently, the conventional approach to the controller design for these systems may not guarantee accurate tracking control performance. An effective neuro-fuzzy controller is proposed to realize an accurate hydraulic servo-system regardless of the uncertainties and the external disturbances. For this purpose, first, we develop a simplified fuzzy logic controller which have multidimensional and unsymmetric membership functions. Secondly, we develop a neural network which consists of the parameters of the fuzzy logic controller. It is show that the neural network has both learning capability and linguistic representation capability. The proposed controller was implemented on a hydraulic servo-system. Feedback error learning architecture is adopted which uses the feedback error directly without passing through the dynamics or inverse transfer function of the hydraulic servo-system to train the neuro-fuzzy controller. A series of simulations was performed for the position-tracking control of the system subjected to external disturbances. The results of simulations show that regardless of inherent non-linearities and disturbances, an accuracy tracking-control performance is obtained using the proposed neuro-fuzzy controller.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.60 no.2
• /
• pp.398-406
• /
• 2011

In this paper, we introduce a new architecture of PSO-based Polynomial Neural Networks (PNN) and discuss its comprehensive design methodology. The conventional PNN is based on a extended Group Method of Data Handling (GMDH) method, and utilized the polynomial order (viz. linear, quadratic, and modified quadratic) as well as the number of node inputs fixed (selected in advance by designer) at Polynomial Neurons located in each layer through a growth process of the network. Moreover it does not guarantee that the conventional PNN generated through learning results in the optimal network architecture. The PSO-based PNN results in a structurally optimized structure and comes with a higher level of flexibility that the one encountered in the conventional PNN. The PSO-based design procedure being applied at each layer of PNN leads to the selection of preferred PNs with specific local characteristics (such as the number of input variables, input variables, and the order of the polynomial) available within the PNN. In the sequel, two general optimization mechanisms of the PSO-based PNN are explored: the structural optimization is realized via PSO whereas in case of the parametric optimization we proceed with a standard least square method-based learning. To evaluate the performance of the PSO-based PNN, the model is experimented with using Gas furnace process data, and pH neutralization process data. For the characteristic analysis of the given entire data with non-linearity and the construction of efficient model, the given entire system data is partitioned into two type such as Division I(Training dataset and Testing dataset) and Division II(Training dataset, Validation dataset, and Testing dataset). A comparative analysis shows that the proposed PSO-based PNN is model with higher accuracy as well as more superb predictive capability than other intelligent models presented previously.

• Journal of Intelligence and Information Systems
• /
• v.9 no.3
• /
• pp.81-97
• /
• 2003

In this paper, we suggest a method of robust watermarking for protection of multimedia data using the wavelet transform and artificial neural network. for the purpose of implementation, we decompose a original image using wavelet transform at level 3. After we classify transformed coefficients of other subbands using neural network except fur the lowest subband LL$_3$, we apply a calculated threshold about chosen cluster as the biggest. We used binary logo watermarks to make sure that it is true or not on behalf of the Gaussian Random Vector. Besides, we tested a method of dual watermark insertion and extraction. For the purpose of implementation, we decompose a original image using wavelet transform at level 3. After we classify transformed coefficients of other subbands using neural network except for the lowest subband LL$_3$, we apply a above mentioned watermark insert method. In the experimental results, we found that it has a good quality and robust about many attacks.