• Journal of Broadcast Engineering
• /
• v.27 no.6
• /
• pp.840-847
• /
• 2022

In a deep neural network (DNN)-based speech enhancement, using global and local input speech information is closely related to model performance. Recently, a nested U-Net structure that utilizes global and local input data information using multi-scale has bee n proposed. This nested U-Net was also applied to speech enhancement and showed outstanding performance. However, a single skip connection used in nested U-Nets must be modified for the nested structure. In this paper, we propose a multi-level skip connection (MLS) to optimize the performance of the nested U-Net-based speech enhancement algorithm. As a result, the proposed MLS showed excellent performance improvement in various objective evaluation metrics compared to the standard skip connection, which means th at the MLS can optimize the performance of the nested U-Net-based speech enhancement algorithm. In addition, the final proposed m odel showed superior performance compared to other DNN-based speech enhancement models.

• Proceedings of the KWS Conference
• /
• 1996.05a
• /
• pp.115-118
• /
• 1996

1. The crack is inifiated at a toe notch and reptured for the skip and continuous welding. 2. Fatigue life of continuous welding is greater than that of skip welding and fatigue life of 2pad continuous welding appears 1.7 times more than that of pad continuous welding. 3. For the skip and continuous welding, fatigue than that of welding materials appears shorter than that of welding materials because of the effect of plastic strain and compressive residual stresses at the crack tip, which are generated by the over loading. 4. Inter-relationship between da/dN-ΔK and delayed phenomenon increases linearly at the state growth area of heat effect section for the skip and continuous welding.

• The Journal of the Korea Contents Association
• /
• v.10 no.8
• /
• pp.38-50
• /
• 2010

In recent years, the number of users who are actively express their opinions about Internet articles is more and more growing up, as the use of cyber community such as weblog or Internet discussion board increases. In fact, it is not difficult to find an article with hundreds of comments in famous Internet discussion boards. Most of the weblogs or Internet discussion boards present comments in the form of list and do not yet support even the basic operation such as searching comments. In this paper, we analysed large sets of comments in Internet discussion board named AGORA. It was found that from the result that the distribution of comment writers follows power-law. So we suppose a new search structure of comments using skip lists. The main idea of our approach is to reflect the probabilistic distribution properties of the commenters following the power-law to the data structure. Our empirical results show that the proposed method performs more efficient in searching the nodes with fewer number of comparison operations than logN, which is the theoretical time complexity of general indexed structure such as B-trees or typical skip lists.

• Journal of Information Processing Systems
• /
• v.19 no.1
• /
• pp.98-108
• /
• 2023

This paper is an attempt to design segmentation method based on fully convolutional networks (FCN) and attention mechanism. The first five layers of the Visual Geometry Group (VGG) 16 network serve as the coding part in the semantic segmentation network structure with the convolutional layer used to replace pooling to reduce loss of image feature extraction information. The up-sampling and deconvolution unit of the FCN is then used as the decoding part in the semantic segmentation network. In the deconvolution process, the skip structure is used to fuse different levels of information and the attention mechanism is incorporated to reduce accuracy loss. Finally, the segmentation results are obtained through pixel layer classification. The results show that our method outperforms the comparison methods in mean pixel accuracy (MPA) and mean intersection over union (MIOU).

• Transactions of the Korean Society of Machine Tool Engineers
• /
• v.10 no.4
• /
• pp.7-13
• /
• 2001

This Study is to investigate the influence of weld residual stresses on the fatigue crack growth behaviors in pressure ves-sel reinforcement. In order to perform this study, the automatically welded specimens are prepared. The material is ASTM A516 grade 60 steel used in pressure vessel mainly. For pad-on-plate of skip welding continuous welding and PWHT specimen, fatigue crack initiation is generally initiat-ed at weld starting and end toe zone, and ruptured at weld starting toe zone, Fatigue life if pad-on-plate continuous speci-men is increased more than that of pad-on-plate skip fillet welding specimene about 85% under low load, about 20% under high load, and decreased than that of two-pad continuous welding specimen about 85%. In da/dN-$\Delta$ Κ curve under low load, pad-on-plate skip fillet welding specimen showed retardation on the initial crack, and the fatigue crack growth rate at the low region of $\Delta$Κ greater specimene E(3.8{\times}10^{-6}$mm/cycle). And the fatigue life of welding specimen was smaller than that of PWHT specimen.

• Journal of Broadcast Engineering
• /
• v.16 no.6
• /
• pp.996-1008
• /
• 2011

This paper proposes a PSNR estimation algorithm of H.264/AVC bitstream for hierarchical B-picture structure. The proposed method consists of a modeling method for DCT coefficients for I-pictures and an error estimation method for blocks encoded by skip mode. The modeling method selects an appropriate model between Laplacian and Cauchy model, and the error of skip mode is estimated from MSE values of reference pictures. Experimental results show that the modeling method predicts more accurate PSNR values than Laplacian and Cauchy model and the error estimation method of skip mode enhances PSNR estimation of hierarchical B-picture structure.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.45 no.3
• /
• pp.86-93
• /
• 2008

The multi-mode terminal modem which is capable of accommodating a variety of wireless communication standards needs versatile arithmetic units for processing a variety of word lengths and wide range of data rates. Since the target hardware is usually designed to meet the required highest performance, it is often wasteful in power consumption especially when low rate data processing cases. Thus, a speed and power adaptability of the arithmetic unit is a desirable feature for the wireless applications. In this paper, we propose a power efficient versatile adder architecture with carry skip logic as a basic building block constructed in hierarchical manner. The validity of the architecture is shown with respect to size, performance, and power efficiency in diverse operating modes.

• Journal of Broadcast Engineering
• /
• v.19 no.2
• /
• pp.257-267
• /
• 2014

Distributed compressive video sensing (DCVS) is a low cost sampling paradigm for video coding based on the compressive sensing and the distributed video coding. In this paper, we propose using a skip-mode coding in DCVS under the assumption that in case of high temporal correlation, temporal interpolation can guarantee sufficiently good quality of nonkey frame, therefore no need to transmit measurement data in such a nonkey frame. Furthermore, we extend it to use a hierarchical structure for better temporal interpolation. Simulation results show that the proposed skip-mode coding can save the average subrate of whole video sequence while the PSNR is reduced only slightly. In addition, by using the proposed scheme, the computational complexity is also highly decreased at decoder on average by 43.75% for video sequences that have strong temporal correlation.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.41 no.6
• /
• pp.99-108
• /
• 2004

In this paper, a new fast detection and clutter rejection method is proposed for CCD-image-based Automatic Target Detection System. For defence application, fast computation is a critical point, thus we concentrated on the ability to detect various targets with simple computation. In training stage, 1D template set is generated by regional vertical projection and K-means clustering, and binary tree structure is adopted to reduce the number of template matching in test stage. We also use adaptive skip-width by Correlation-based Adaptive Predictive Search(CAPS) to further improve the detecting speed. In clutter rejection stage, we obtain Fourier Descriptor coefficients from boundary information, which are useful to rejected clutters.

• The Journal of Korea Robotics Society
• /
• v.13 no.1
• /
• pp.8-15
• /
• 2018

This paper proposes a convolutional neural network model for distinguishing areas occupied by obstacles from a LiDAR image converted from a 3D point cloud. The channels of a LiDAR image used as input consist of the distances to 3D points, the reflectivities of 3D points, and the heights of 3D points from the ground. The proposed model uses a LiDAR image as an input and outputs a result of a segmented LiDAR image. The proposed model adopts refinement modules with skip connections to segment a LiDAR image. The refinement modules with skip connections in the proposed model make it possible to construct a complex structure with a small number of parameters than a convolutional neural network model with a linear structure. Using the proposed model, it is possible to distinguish areas in a LiDAR image occupied by obstacles such as vehicles, pedestrians, and bicyclists. The proposed model can be applied to recognize surrounding obstacles and to search for safe paths.