• Journal of Digital Convergence
• /
• v.11 no.11
• /
• pp.381-387
• /
• 2013

Recently, emotional contents that induce emotions and respond to emotions are given attention in the field of cultural industries, and extracting emotion caused by multimedia contents is being noted. Furthermore, since multimedia contents have been quickly produced and distributed these days, researches automatically to extract the feeling of multimedia contents are being accelerated. In this paper, we will study the method of emotional value extraction in the multimedia contents using the volume value of the multimedia contents in a certain frequency among sound informations. This study allows to extract the emotion of multimedia contents automatically, and the extracted information will be used to provide user's current emotion, weather, etc. for the users.

• 한국정보컨버전스학회:학술대회논문집
• /
• 2008.06a
• /
• pp.153-156
• /
• 2008

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2010.07a
• /
• pp.511-512
• /
• 2010

• Journal of Korea Multimedia Society
• /
• v.21 no.2
• /
• pp.232-241
• /
• 2018

Operational Transformation (OT) algorithms for real-time collaborative editing systems are becoming increasingly important due to the increased demand for collaborative data processing. The operational transformation algorithm is a technique for real-time concurrency control and consistency maintenance with non-locking technique, and many studies have been conducted to overcome three issues of convergence, causality-prevention, and intention-prevention. However, previous work has the disadvantage of wasting memory by storing all operations that occurred during an edit operation in the history buffer to solve this problem. Therefore, we propose a memory-saving real-time collaborative editing system that maintains a constant memory space and concurrency control through a method of applying the valid-time to each user-generated operation in order to reduce memory waste. This system prevents long-term memory occupation of client-generated operations, thus it reduces the space and time complexity even with low-rate of collaboration work, so that the performance degradation avoids.

• Journal of Korea Multimedia Society
• /
• v.12 no.9
• /
• pp.1210-1216
• /
• 2009

In this paper we analyze the convergence behavior of the recursive least rank (RLR) L-filter. The RLR L-filter is an order statistics filter, filter coefficients of which are the weights according to the order of magnitude of inputs. And RLR L-filter is a non-linear adaptive filter, that uses RLR algorithm for coefficient updating. The RLR algorithm is a non-linear adaptive algorithm based on rank estimates in Robust statistics. The mean and mean-squared convergence behavior of the RLR L-filter is examined with variable step-sizes. The RLR L-filter adapts the median filter type to the heavy-tailed distribution function of impulse noise, and adapts the average filter type to Gaussian noises.

• Journal of Korea Multimedia Society
• /
• v.21 no.10
• /
• pp.1150-1161
• /
• 2018

Classification is widely used in medical images to categorize patients and non-patients. However, conventional classification requires a complex procedure, including some rigid steps such as pre-processing, segmentation, feature extraction, detection, and classification. In this paper, we propose a novel convolutional neural network (CNN), called LeukemiaNet, to specifically classify two different types of leukemia, including acute lymphoblastic leukemia (ALL) and acute myeloid leukemia (AML), and non-cancerous patients. To extend the limited dataset, a PCA color augmentation process is utilized before images are input into the LeukemiaNet. This augmentation method enhances the accuracy of our proposed CNN architecture from 96.9% to 97.2% for distinguishing ALL, AML, and normal cell images.

• Journal of Korea Multimedia Society
• /
• v.22 no.11
• /
• pp.1269-1279
• /
• 2019

In medical image analysis, segmentation is considered as a vital process since it partitions an image into coherent parts and extracts interesting objects from the image. In this paper, we consider automatic segmentations of OCT retinal images to find six layer boundaries using convolutional neural networks. Segmenting retinal images by layer boundaries is very important in diagnosing and predicting progress of eye diseases including diabetic retinopathy, glaucoma, and AMD (age-related macular degeneration). We applied well-known CNN architecture for general image segmentation, called Segnet, U-net, and CNN-S into this problem. We also proposed a shortest path-based algorithm for finding the layer boundaries from the outputs of Segnet and U-net. We analysed their performance on public OCT image data set. The experimental results show that the Segnet combined with the proposed shortest path-based boundary finding algorithm outperforms other two networks.

• Journal of Multimedia Information System
• /
• v.4 no.4
• /
• pp.311-316
• /
• 2017

High Efficiency Video Coding (HEVC) is the newest video coding standard for improvement in video data compression. This new standard provides a significant improvement in picture quality, especially for high-resolution videos. A quadtree-based structure is created for the encoding and decoding processes and the rate-distortion (RD) cost is calculated for all possible dimensions of coding units in the quadtree. To get the best combination of the block an optimization process is performed in the encoder, called rate distortion optimization (RDO). In this work we are proposing a novel approach to enhance the overall RDO process of HEVC encoder. The proposed algorithm is performed in two steps. In the first step, like HEVC, it performs general rate distortion optimization. The second step is an extra checking where a SSIM based cost is evaluated. Moreover, a fast SSIM (FSSIM) calculation technique is also proposed in this paper.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.12
• /
• pp.6152-6168
• /
• 2017

Recently, the biometric-based authentication systems such as FIDO (Fast Identity Online) are increased in mobile computing environments. The biometric-based authentication systems are performed on the mobile devices with the battery, the improving energy efficiency is important issue. In the case, the size of images (i.e., face, fingerprint, iris, and etc.) affects both recognition accuracy and energy consumption, and hence the tradeoff analysis between the both recognition accuracy and energy consumption is necessary. In this paper, we propose an energy-efficient way to authenticate based on biometric information with tradeoff analysis between the both recognition accuracy and energy consumption in multimedia IoT (Internet of Things) transmission environments. We select the facial information among biometric information, and especially consider the multicore-based mobile devices. Based on our experimental results, we prove that the proposed approach can enhance the energy efficiency of GABOR+LBP+GRAY VALUE, GABOR+LBP, GABOR, and LBP by factors of 6.8, 3.6, 3.6, and 2.4 over the baseline, respectively, while satisfying user's face recognition accuracy.

• Journal of Korea Multimedia Society
• /
• v.25 no.11
• /
• pp.1643-1652
• /
• 2022

Most of the existing machine reading research has used Recurrent Neural Network (RNN) and Convolutional Neural Network (CNN) algorithms as networks. Among them, RNN was slow in training, and Question Answering Network (QANet) was announced to improve training speed. QANet is a model composed of CNN and self-attention. CNN extracts semantic and syntactic information well from the local corpus, but there is a limit to extracting the corresponding information from the global corpus. Graph Convolutional Networks (GCN) extracts semantic and syntactic information relatively well from the global corpus. In this paper, to take advantage of this strength of GCN, we propose I-QANet, which changed the CNN of QANet to GCN. The proposed model performed 1.2 times faster than the baseline in the Stanford Question Answering Dataset (SQuAD) dataset and showed 0.2% higher performance in Exact Match (EM) and 0.7% higher in F1. Furthermore, in the Korean Question Answering Dataset (KorQuAD) dataset consisting only of Korean, the learning time was 1.1 times faster than the baseline, and the EM and F1 performance were also 0.9% and 0.7% higher, respectively.