• Journal of Korea Multimedia Society
• /
• v.24 no.11
• /
• pp.1552-1559
• /
• 2021

Convolution neural networks (CNNs) show notable performance in image processing and are used as representative core models. CNNs extract and learn features from large amounts of train dataset. In general, it has a structure in which a convolution layer and a fully connected layer are stacked. The core of CNN is the convolution layer. The size of the kernel used for feature extraction and the number that affect the depth of the feature map determine the amount of weight parameters of the CNN that can be learned. These parameters are the main causes of increasing the computational complexity and memory usage of the entire neural network. The most computationally expensive components in CNNs are fully connected and spatial convolution computations. In this paper, we propose a Fourier Convolution Neural Network that performs the operation of the convolution layer in the Fourier domain. We work on modifying and improving the amount of computation by applying the fast fourier transform method. Using the MNIST dataset, the performance was similar to that of the general CNN in terms of accuracy. In terms of operation speed, 7.2% faster operation speed was achieved. An average of 19% faster speed was achieved in experiments using 1024x1024 images and various sizes of kernels.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2012.05a
• /
• pp.646-648
• /
• 2012

In this paper, an image detection technique is proposed to extract image features in a mobile terminal environment. To detect objects, the HSI color model of the image is used. The object's corner points are detected using the Harris corner detection method. Finally we detect the object of interest using region growing The experiment results show that the proposed method improves detection performance and reduces the amount of computation.

• Proceedings of the IEEK Conference
• /
• 2000.11d
• /
• pp.199-202
• /
• 2000

Speaking Rate has variety depends on the situation and habit of speakers. It has been many studied about speaking rate In speaker recognition. The study of speaking rate in speech recognition is one of considerable matter when It is recognized the speakers and it is measured by many speech data base and complicate estimation for accuracy. In this paper, conventional vocoder process the speech signal when encoding and transmitting without regard to speaking rate so in order to apply the speaking rate for vocoder It should be considered the simpler algorithm and less computation amount than the conventional method of speaking rate used In speech recognition. We proposed the speaking rate algorithm which is used the simple parameter with Line Spectrum Pair (LSP). The proposed peaking rate method is measured by the information of LSP in speech. We measured the variety rate of phenomenon about utterances which have different velocity, respectively. As a result, It has distinct variation rate of phenomenon between utterances uttered fast and slow and the rate is 42.8% higher in case of uttered fast than in case of uttered slow.

• Journal of the Korean Institute of Telematics and Electronics A
• /
• v.29A no.11
• /
• pp.70-77
• /
• 1992

Generally, the conventional tracking algorithms are very limited in the practical applications because of that the computation load is exponentially increased as the number of targets being tracked is increase. Recently, to overcome this kind of limitation, some new tracking methods based on neural network algorithms which have learning and parallel processing capabilities are introduced. By application of neural networks to multi-target tracking problems, the tracking system can be made computationally independent of the number of objects being tracked, through their characteristics of massive parallelism and dense interconnectivity. In this paper, a new neural network tracking algorithm, which has capability of adaptive target tracking with little increase of the amount of calculation under the clutter and noisy environments, is suggested and the possibility of real-time multi-target tracking system based on neural networks is also demonstrated through some good computer simulation results.

• Proceedings of the IEEK Conference
• /
• 2005.11a
• /
• pp.987-990
• /
• 2005

An important issue in the field of face recognitions and man-machine interfaces is an automatic detection of faces in visual scenes. it should be computationally fast enough to allow an online detection. In this paper we describe our ongoing work on face detection that models the face appearance by edge orientation and color distribution. We show that edge orientation is a powerful feature to describe objects like faces. We present a method for face region detection using edge orientation and a method for face feature detection using local color information. We demonstrate the capability of our detection method on an image database of 1877 images taken from more than 700 people. The variations in head size, lighting and background are considerable, and all images are taken using low-end cameras. Experimental results show that the proposed scheme achieves 94% detection rate with a resonable amount of computation time.

• Proceedings of the Korean Society for Technology of Plasticity Conference
• /
• 2001.10a
• /
• pp.81-84
• /
• 2001

An inverse finite element approach is employed for more capability to design the optimum blank shape from the desired final shape with small amount of computation time and effort. In some drawing or stamping simulation with inverse method, it is difficult to apply inverse scheme due to the large aspect ratio or steep vertical angle of inclination. The reason is that initial guesses are hard to make out with present method for those cases. In this paper, a direct mesh marring scheme to generate initial guess on the sliding constraint surface described by finite element patches is suggested for one step inverse analysis to calculate initial blank shape. Radial type mapping is adopted for the simulation of rectangular cup drawing process with large aspect ratio and parallel type mapping for the simulation of S-Rail forming process with steep vertical angle of inclination.

• Proceedings of the KSME Conference
• /
• 2001.06c
• /
• pp.580-585
• /
• 2001

A response surface method(RSM) is utilized for structural optimization and implemented on a parametric CAD platform. Once an approximation of the performance function is made, no formal design sensitivity analysis is necessary. The approximation gives the designer the sensitivity information and furthermore intuition on the performance functions. The scheme for the design of experiment chosen for the RSM has a large influence on the accuracy of converged solutions and the amount of computation. The D-optimal design criterion as implemented in this paper is found efficient for the structural optimization. The program is developed on a parametric CAD platform and tested using several shape design problems of such as a torque arm and a belt clip. It is observed that the RSM used provides a faster convergence than other approximation methods for design sensitivity.

• Proceedings of the KSME Conference
• /
• 2004.11a
• /
• pp.522-527
• /
• 2004

Resealed Simulated Annealing (RSA) has been devised for improving the disadvantage of Simulated Annealing (SA) which require tremendous amount of computation time. RSA and SA have been for optimization of satellite structures and for comparison of results from two algorithms. As a practical application, a satellite structure is optimized by the two algorithms. Weights of satellite upper platform and propulsion module are minimized. MSC/NASTRAN is used for the static and dynamic analysis. The optimization results of the RSA are compared with results of the classical SA. The numbers of optimization iterations could be effectively reduced by the RSA.

• Transactions of the Korean Society of Automotive Engineers
• /
• v.6 no.5
• /
• pp.45-54
• /
• 1998

For proper release of side airbags, the onset of crash should be detected first. After crash detection, the algorithm has to make a decision whether the side airbag deployment is necessary. If the deployment is necessary, proper timing has to be provided for the maximum protection of driver or passenger. The side airbag release algorithm should be robust against the statistical deviations which are inherent to experimental crash test data. Deterministic optimization algorithms cannot be used for the side aribag release algorithm since the objective function cannot be expressed in a closed form. From this background, genetic algorithm has been used for the optimization. The optimization requires moderate amount of computation and gives satisfactory results.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2005.06a
• /
• pp.1988-1991
• /
• 2005

A micro fluid actuator driven by electromagnetic force at MEMS(Micro Electro Mechanical System) level has been designed. The operation of the actuator was simulated in three steps. First, fluid flow analysis has been performed to determine the actuator load. With the load, dynamic behavior of the actuator structure has been analysed. Finally, fluid-structure interaction analysis has been performed to predict the performance of the actuator. To avoid excessive amount of computation, axisymmetric and plane strain 2-D models were used.