• 한국멀티미디어학회논문지
• /
• 제24권7호
• /
• pp.880-889
• /
• 2021

In everyday life, recognizing people's emotions from their frames is essential and is a popular research domain in the area of computer vision. Visual emotion has a severe class imbalance in which most of the data are distributed in specific categories. The existing methods do not consider class imbalance and used accuracy as the performance metric, which is not suitable for evaluating the performance of the imbalanced dataset. Therefore, we proposed a method for recognizing visual emotion using balanced data augmentation to address the class imbalance. The proposed method generates a balanced dataset by adopting the random over-sampling and image transformation methods. Also, the proposed method uses the Focal loss as a loss function, which can mitigate the class imbalance by down weighting the well-classified samples. EfficientNet, which is the state-of-the-art method for image classification is used to recognize visual emotion. We compare the performance of the proposed method with that of conventional methods by using a public dataset. The experimental results show that the proposed method increases the F1 score by 40% compared with the method without data augmentation, mitigating class imbalance without loss of classification accuracy.

• 한국멀티미디어학회논문지
• /
• 제25권6호
• /
• pp.886-893
• /
• 2022

Outlier detection refers to the task of detecting data that deviate significantly from the normal data distribution. Most outlier detection methods compute an outlier score which indicates the degree to which a data sample deviates from normal. However, setting a threshold for an outlier score to determine if a data sample is outlier or normal is not trivial. In this paper, we propose a binary prediction method for outlier detection based on spectral clustering and one-class SVM ensemble. Given training data consisting of normal data samples, a clustering method is performed to find clusters in the training data, and the ensemble of one-class SVM models trained on each cluster finds the boundaries of the normal data. We show how to obtain a threshold for transforming outlier scores computed from the ensemble of one-class SVM models into binary predictive values. Experimental results with high dimensional text data show that the proposed method can be effectively applied to high dimensional data, especially when the normal training data consists of different shapes and densities of clusters.

• 정보처리학회논문지B
• /
• 제8B권5호
• /
• pp.501-506
• /
• 2001

최근에 인터넷과 통신망의 활성화로 인하여 멀티미디어 정보들을 효율적으로 관리하고 서비스하기 위한 여러 가지 방법들의 제안되고 있다. 본 논문에서는 퍼지 기반의 멀티미디어 사서함 구축을 위한 객체관리기로서 $\alpha$-cut 을 이용한 FBOM을 제안한다. 제안된 시스템은 퍼지 필터링을 이용하여 객체들을 고나리하기 위해 객체 분류, 퍼지 필터링, 클래스 생성구조를 이용한다. 또한 제안된 시스템의 성능을 알아보기 위해 1000개의 멀티미디어 정보를 이용하여 실험을 수행하고, 랜덤 키 방법과 FBOM 방법을 비교 분석한다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제11권8호
• /
• pp.3904-3917
• /
• 2017

IP multimedia subsystem (IMS) is an open standardized architecture for delivering multimedia service over IP network in a route-agnostic manner. With the increasing popularity of conversational class service, the delivery of a traffic flow with a certain bandwidth demand over a single network path is either not possible or not cost-effective. Multipath transmission is considered to be a promising solution to provide high-quality delivery service. This paper proposes a software defined service overlay network (SDSON) based multipath transmission framework for IMS, which is complementary to existing network architecture. The framework transforms original two-party session negotiation into three-party session negotiation that supports participants to negotiate multipath transmission capacity and path information by signaling message. Based on existing IETF standards, SIP and SDP are scalable to support these functions. Finally, the proposed framework is fully implemented on open source platform and examined by experiments. Experimental results show that multipath-enabled IMS is an effective way to improve the delivery performance of conversational class service.

• 한국컴퓨터산업학회논문지
• /
• 제7권1호
• /
• pp.1-6
• /
• 2006

기존의 객체지향 멀티미디어 베이스에 경험적 분류 모델에 기반을 둔 새로운 클래스를 도입한 확장된 객체 지향 멀티미디어 베이스의 모델을 설계한다. 이를 구현하기 위해 기존의 객체 멀티미디어 베이스에 경험적 분류 클래스를 도입하였으며, 이 클래스들을 연산하기 위한 설계된 객체 지향 멀티미디어 프로그램을 설계하였다. 그리고 설계된 객체 지향 멀티미디어 프로그램을 이용하여 비교집단과 실험처치된 실험집단과 비교하여 창의성과의 상관관계에 대해 연구한다.

• 한국통신학회논문지
• /
• 제23권9A호
• /
• pp.2249-2257
• /
• 1998

In this paper, we propose the enhanced priority control algorithm that adaptively controls the cell service ratio according to the relative cell occupancy ratio of buffer. The asynchronous transfer mode (ATM) provides the means to support various multimedia services in broadband networks. To support multimedia services, various data traffics of different priorities should be controlled effectively. And also it needs congestion control functions required in the netowrk to carry out the control operation. To accomplish this in a flexible and effective manner, priority classes for the different services ar ecommonly used. The proposed enhanced priority control mechanism have two service calsses of the delay sensitive class and the loss sensitive class. The simulation results show that te proposed control mechanism improves the QoS, the charateristics of cell loss probability and mean cell delay time, by selecting propeor relativ ecell occupancy ratio of buffer and the average arrival rate.

• 한국통신학회논문지
• /
• 제34권3B호
• /
• pp.288-303
• /
• 2009

In the near future it is foreseen that a genuine multimedia service over the WiMAX system is provided in a worldwide manner by exploiting the QoS technologies introduced in the wireless and wired broadband network. In this work we propose a pricing scheme for the multimedia service over the generic WiMAX system that supports a full QoS functionality. We assume real-time services such as the voice and video as well as the nonreal-time service such as the conventional high-speed data, and we propose a pricing and charging scheme for those services by investigating the inherent characteristics of those services and the multiple-class of QoS-service provided to them. After that we propose a method to compute expected revenue that is obtained from the WiMAX system by using an analytic method to estimate the usage of the bandwidth resources for the different class of services. Via numerical experiment, we verify the implication of the work.

• 한국컴퓨터정보학회논문지
• /
• 제23권1호
• /
• pp.139-145
• /
• 2018

This paper recommends flipped learning as a method to improve the learning abilities and the level of software utilization when it comes to using computers in children education institutes. Flipped learning enables a class fully making use of the up-to-date multimedia-related technology. Especially, flipped learning leads a participation-oriented class rather than lecture-based ones. Young child's teachers can, not only improve their capabilities to utilize multimedia, but also manage classes that follow the trend of the fourth industrial revolution. Therefore, this paper introduces the importance of media education when it comes to training preliminary teachers and suggests a flipped learning curriculum. This paper finds significance in future efficient education for raising creative and integrated thinking children.

• 한국멀티미디어학회논문지
• /
• 제10권11호
• /
• pp.1472-1482
• /
• 2007

로더/링커는 자바 클래스 파일 또는 .NET 환경의 중간 표현인 PE 파일을 입력으로 받아 검증, 레졸루션, 초기화, 실행에 필요한 최적화된 정보 저장 등 실질적인 실행에 필요한 모든 정보 생성 및 무결성을 보장하는 아주 중요한 부분이다. 본 논문에서는 자바 클래스 파일과 .NET 환경의 PE 파일에 대한 통합 로더/링커 시스템을 개발하고자 한다. 이를 위해, 자바 클래스 파일과 .NET PE 파일 정보를 모두 저장할 수 있는 새로운 실행 파일 포맷(*.evm) 및 메모리 포맷을 설계했으며 저장된 실행 정보를 활용하여 JVM 또는 .NET 환경에서 실행할 수 있도록 링커/로더 시스템을 구현하였다.

• 한국멀티미디어학회논문지
• /
• 제22권6호
• /
• pp.655-664
• /
• 2019

In this paper, we propose a novel way of combining multiple deep convolutional neural network (DCNN) architectures which work well for accurate video face identification by adopting a serial combination of 3D and 2D DCNNs. The proposed method first divides an input video sequence (to be recognized) into a number of sub-video sequences. The resulting sub-video sequences are used as input to the 3D DCNN so as to obtain the class-confidence scores for a given input video sequence by considering both temporal and spatial face feature characteristics of input video sequence. The class-confidence scores obtained from corresponding sub-video sequences is combined by forming our proposed class-confidence matrix. The resulting class-confidence matrix is then used as an input for learning 2D DCNN learning which is serially linked to 3D DCNN. Finally, fine-tuned, serially combined DCNN framework is applied for recognizing the identity present in a given test video sequence. To verify the effectiveness of our proposed method, extensive and comparative experiments have been conducted to evaluate our method on COX face databases with their standard face identification protocols. Experimental results showed that our method can achieve better or comparable identification rate compared to other state-of-the-art video FR methods.