• Journal of Information Processing Systems
• /
• v.14 no.1
• /
• pp.176-190
• /
• 2018

This paper presents a novel convolutional neural network based multi-feature fusion learning method for non-rigid 3D model retrieval, which can investigate the useful discriminative information of the heat kernel signature (HKS) descriptor and the wave kernel signature (WKS) descriptor. At first, we compute the 2D shape distributions of the two kinds of descriptors to represent the 3D model and use them as the input to the networks. Then we construct two convolutional neural networks for the HKS distribution and the WKS distribution separately, and use the multi-feature fusion layer to connect them. The fusion layer not only can exploit more discriminative characteristics of the two descriptors, but also can complement the correlated information between the two kinds of descriptors. Furthermore, to further improve the performance of the description ability, the cross-connected layer is built to combine the low-level features with high-level features. Extensive experiments have validated the effectiveness of the designed multi-feature fusion learning method.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.54 no.3
• /
• pp.156-165
• /
• 2005

Human being recognizes the physical world by integrating a great variety of sensory inputs, the information acquired by their own action, and their knowledge of the world using hierarchically parallel-distributed mechanism. In this paper, authors propose the sensor fusion system that can recognize multiple 3D objects from 2D projection images and tactile informations. The proposed system focuses on improving recognition performance of 3D objects. Unlike the conventional object recognition system that uses image sensor alone, the proposed method uses tactual sensors in addition to visual sensor. Neural network is used to fuse the two sensory signals. Tactual signals are obtained from the reaction force of the pressure sensors at the fingertips when unknown objects are grasped by four-fingered robot hand. The experiment evaluates the recognition rate and the number of learning iterations of various objects. The merits of the proposed systems are not only the high performance of the learning ability but also the reliability of the system with tactual information for recognizing various objects even though the visual sensory signals get defects. The experimental results show that the proposed system can improve recognition rate and reduce teeming time. These results verify the effectiveness of the proposed sensor fusion system as recognition scheme for 3D objects.

• Journal of the Korean Geophysical Society
• /
• v.9 no.2
• /
• pp.99-103
• /
• 2006

3D object recognition independent of translation and rotation using an ultrasonic sensor array, invariant moment vectors and SOFM(Self Organizing Feature Map) neural networks is presented. Using invariant moment vectors of the acquired 16×8 pixel data of square, rectangular, cylindric and regular triangular blocks, 3D objects could be classified by SOFM neural networks. Invariant moment vectors are constant independent of translation and rotation. The recognition rates for the training and testing data were 95.91% and 92.13%, respectively.

• Proceedings of the IEEK Conference
• /
• 1999.06a
• /
• pp.496-499
• /
• 1999

This paper describes a medular neural network(MNN) for a vision system which tracks a given object using a sequence of images from a camera unit. The MNN is used to precisely recognize the given robot endeffector and to minize the processing time. Since the robot endeffector can be viewed in many different shapes in 3-D space, a MNN structure, which contains a set of feedforwared neural networks, co be more attractive in recognizing the given object. Each single neural network learns the endeffector with a cluster of training patterns. The training patterns for a neural network share the similar charateristics so that they can be easily trained. The trained MNN is less sensitive to noise and it shows the better performance in recognizing the endeffector. The recognition rate of MNN is enhanced by 14% over the single neural network. A vision system with the MNN can precisely recognize the endeffector and place it at the center of a display for a remote operator.

• Journal of Power Electronics
• /
• v.12 no.3
• /
• pp.468-476
• /
• 2012

The interior permanent magnet synchronous motor (IPMSM) has been widely used in electric vehicle applications due to its excellent power to weigh ratio. This paper proposes the maximum torque control of an IPMSM drive using an adaptive learning (AL) fuzzy neural network (FNN) and an artificial neural network (ANN). This control method is applicable over the entire speed range while taking into consideration the limits of the inverter's rated current and voltage. This maximum torque control is an executed control through an optimal d-axis current that is calculated according to the operating conditions. This paper proposes a novel technique for the high performance speed control of an IPMSM using AL-FNN and ANN. The AL-FNN is a control algorithm that is a combination of adaptive control and a FNN. This control algorithm has a powerful numerical processing capability and a high adaptability. In addition, this paper proposes the speed control of an IPMSM using an AL-FNN, the estimation of speed using an ANN and a maximum torque control using the optimal d-axis current according to the operating conditions. The proposed control algorithm is applied to an IPMSM drive system. This paper demonstrates the validity of the proposed algorithms through result analysis based on experiments under various operating conditions.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.3 no.1
• /
• pp.112-120
• /
• 2003

This paper reports the development work carried out in respect of a proposed application of Neural Network approach for the Korean Next generation Reactor (KNGR) now referred as APR-1400. The emphasis is on establishing the methodology and the approach to be adopted towards realizing this application in the next generation reactors. Keeping in view the advantages and limitation of Artificial Neural Network Approach, the role of ANN has been limited to plant status or to be more precise plant transient monitoring. The simulation work carried out so far and the results obtained shows that artificial neural network approach caters to the requirements of plant status monitoring and qualifies to be incorporated as a part of proposed operator support systems of the referenced nuclear power plant.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.3
• /
• pp.228-233
• /
• 2021

In this paper, we propose a method to extract and classify features of multi-channel ElectroEncephalo Graphy (EEG) that change according to various musical tempo stimuli. In the proposed method, a 3D convolutional bidirectional gated recurrent neural network extracts spatio-temporal and long time-dependent features from the 3D EEG input representation transformed through the preprocessing. The experimental results show that the proposed tempo stimuli classification method is superior to the existing method and the possibility of constructing a music-based brain-computer interface.

• Journal of the Korea Society of Computer and Information
• /
• v.24 no.6
• /
• pp.11-19
• /
• 2019

In this paper, a new Scene-based Nonuniformity Correction (SBNUC) method is proposed by applying Image Roughness-like and Spatial Noise cost functions on deep neural network structure. The classic approaches for nonuniformity correction require generally plenty of sequential image data sets to acquire accurate image correction offset coefficients. The proposed method, however, is able to estimate offset from only a couple of images powered by the characteristic of deep neural network scheme. The real world SWIR image set is applied to verify the performance of proposed method and the result shows that image quality improvement of PSNR 70.3dB (maximum) is achieved. This is about 8.0dB more than the improved IRLMS algorithm which preliminarily requires precise image registration process on consecutive image frames.

• Journal of Information Processing Systems
• /
• v.15 no.5
• /
• pp.1179-1191
• /
• 2019

We design an ingenious view-pooling method named learning-based multiple pooling fusion (LMPF), and apply it to multi-view convolutional neural network (MVCNN) for 3D model classification or retrieval. By this means, multi-view feature maps projected from a 3D model can be compiled as a simple and effective feature descriptor. The LMPF method fuses the max pooling method and the mean pooling method by learning a set of optimal weights. Compared with the hand-crafted approaches such as max pooling and mean pooling, the LMPF method can decrease the information loss effectively because of its "learning" ability. Experiments on ModelNet40 dataset and McGill dataset are presented and the results verify that LMPF can outperform those previous methods to a great extent.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.27 no.7
• /
• pp.1110-1119
• /
• 2003

This paper deals with the synthesis of the 3-dimensional grasp planning for unknown objects. Previous studies have many problems, which the estimation time for finding the grasping points is much long and the analysis used the not-perfect 3-dimensional modeling. To overcome these limitations in this paper new algorithm is proposed, which algorithm is achieved by two steps. First step is to find the whole 3-dimensional geometrical modeling for unknown objects by using stereo matching. Second step is to find the optimal grasping points for unknown objects by using the neural network trained by the result of optimization using genetic algorithm. The algorithm is verified by computer simulation, comparing the result between neural network and optimization.