• 한국통신학회논문지
• /
• 제19권10호
• /
• pp.2028-2042
• /
• 1994

ART(Adaptive Resonance Theory) 신경회로망과 같은 자기조직망에서 신호와 잡음을 적절히 정의한다는 것은 어려운 문제이다. 즉, 한 입력 패턴의 일부분이 어떤 패턴에서는 입력 패턴의 신호로 다루어지나 다른 패턴에서는 잡음으로 취급되어야 할 대도 있다. ART 신경회로망 모델은 신호와 잡음의 정의를 문맥과 학습에 따라 적절하게 규정하기 위하여 계산 단위를 자동적으로 자기척도(Self-Scaling 할 수 있는 기능을 가지고 있다. ART 모델에서의 이러한 자기 척도 기능은 입력 패턴들이 유사한 성질을 가진 경우에는 유효하게 잘 동작한다. 그러나 ART 모델은 기본적으로 하나의 경계 인수에 의해 패턴을 분류하기 때문에 여러가지 성질이 복합된 입력 패턴을 효율적으로 분류하기가 어렵다. 예를 들어 패턴들을 자세하게 분류하기 위하여 경계 인수의 값을 크게 하면 잡음으로 취급되어야 할 부분이 신호로 취급되어 불필요한 인식 부류가 발생한다. 또한 경계 인수를 작게 하면 패턴을 구별하기 위한 중요한 정보가 잡음으로 취급되는 경우가 발생하여 비효율적인 분류를 한다. 본 논문에서는 ART 모델의 이러한 문제점을 해결하기 위하여 복합 특징을 분리 처리할 수 잇는 모듈화된 Coupled-ART 신경회로망 모델을 제안하였다. Coupled-ART 신경회로망 모델은 신경회로망의 구조를 기능별로 모듈화하고 이러한 모듈들을 서로 밀착된 구조로 결합하여 확장된 기능을 수행하는 형태로 구성하였다. 이러한 모듈화된 신경회로망을 통해 패턴 인식 과정에서 다양한 크기나 성질을 가진 특징들에 대한 분류를 비슷한 크기나 성질을 가진 특징들끼리 분리하여 분류를 하였다. 그리고 본 논문에서 제안한 상위층에서 각 모듈의 처리 결과를 종합하여 최종적인 분류를 함으로써 기존의 ART 모델보다 더 효율적으로 패턴을 분류할 수 있다.28.8%$)에서 높고 60 및 40%수분구(水分區)($23.6{\sim}24.1%$)에서 낮은 편이었다. 그러나 옥수수의 조섬유함량(粗纖維含量)에 따라 큰 차이(差異)가 없었다. 건엽(乾葉)의 조단백질함량(粗蛋白質含量)에 따라 큰 차이(差異)가 없었다. 건엽(乾葉)의 조단백질함량(粗蛋白質含量)은 60%수분구(水分區)($14.2{\sim}21.6%$) 및 40%수분구(水分區)($13.8{\sim}16.0%$)가 다른 고토양수분구(高土壤水分區)($7.3{\sim}13.9%$)보다 높은 편이었다. 5. 건경중(乾莖中)의 조섬유함량(粗纖維含量)은 $24.6{\sim}36.7%$로서 건엽중(乾葉中)의 함량(含量)보다 월등히 높았고 조단백질함량(粗蛋白質含量)은 $2.0{\sim}5.3%$로서 건엽중(乾葉中)의 함량(含量)보다 현저히 낮았다. 특(特)히 P.931의 건경중(乾莖中)의 조섬유함량(粗纖維含量)은 다른 작물(作物)에 비해 현저(顯著)히 높은 편이었다.적차이(量的差異)를 나타냈다.間)에는 부(負)(-)의 상관(相關)이 있다.($P{\leq}0.01%$). 5. NEL 및 starch value 환경온도(環境溫度)가 상승(上昇)됨에 따라 감소(減少)된다. 4 엽기(葉期) sorghum식물(植物)의 환경온도(環境溫度)를 달리 하였을 때 NEL가치(價値)는 각각(各各) 4.87MJ($30/25^{\circ}C$), 5.46MJ($25/20^{\circ}C$) 및 5.81MJ/kg($18/8^{\circ}C$)로 변(變)하여 고온(高溫)에서 net energy lactation 축적(蓄積)이 크게 감소(減少)되었다.다.

• 한국정보과학회논문지:소프트웨어및응용
• /
• 제35권8호
• /
• pp.463-478
• /
• 2008

본 논문에서는 유전자 사이의 상관계수가 높은 마이크로어레이 데이타에 대하여 제안하는 알고리즘을 통해 상관계수가 낮은 유전자들의 부집합을 만들고, 이에 대해 적합 함수를 통한 평가로 기존 방법론이 가지는 한계를 극복할 수 있도록 하였다. 기존 방법론은 개별 특징의 평가를 통해 중복 특징을 제거하며, 상관계수에 대한 고려가 없어 선택된 유전자 부집합들의 상관계수가 논은 문제가 있었다. 이에 따라 제안하는 알고리즘은 특징간의 관계를 평가하는 Feature Wrapping 기법을 활용하여, 추출된 유전자 부집합에 포함된 유전자 사이의 상관관계가 낮고, 클래스 구분력이 높은 특징을 갖도록 하였다.

• International Journal of Contents
• /
• 제15권4호
• /
• pp.59-64
• /
• 2019

In recent years, emotion recognition has been an interesting and challenging topic. Compared to facial expressions and speech modality, gesture-based emotion recognition has not received much attention with only a few efforts using traditional hand-crafted methods. These approaches require major computational costs and do not offer many opportunities for improvement as most of the science community is conducting their research based on the deep learning technique. In this paper, we propose an end-to-end deep learning approach for classifying emotions based on bodily gestures. In particular, the informative keyframes are first extracted from raw videos as input for the 3D-CNN deep network. The 3D-CNN exploits the short-term spatiotemporal information of gesture features from selected keyframes, and the convolutional LSTM networks learn the long-term feature from the features results of 3D-CNN. The experimental results on the FABO dataset exceed most of the traditional methods results and achieve state-of-the-art results for the deep learning-based technique for gesture-based emotion recognition.

• International Journal of Computer Science & Network Security
• /
• 제21권8호
• /
• pp.288-296
• /
• 2021

Automated face recognition in a runtime environment is gaining more and more important in the fields of surveillance and urban security. This is a difficult task keeping in mind the constantly volatile image landscape with varying features and attributes. For a system to be beneficial in industrial settings, it is pertinent that its efficiency isn't compromised when running on roads, intersections, and busy streets. However, recognition in such uncontrolled circumstances is a major problem in real-life applications. In this paper, the main problem of face recognition in which full face is not visible (Occlusion). This is a common occurrence as any person can change his features by wearing a scarf, sunglass or by merely growing a mustache or beard. Such types of discrepancies in facial appearance are frequently stumbled upon in an uncontrolled circumstance and possibly will be a reason to the security systems which are based upon face recognition. These types of variations are very common in a real-life environment. It has been analyzed that it has been studied less in literature but now researchers have a major focus on this type of variation. Existing state-of-the-art techniques suffer from several limitations. Most significant amongst them are low level of usability and poor response time in case of any calamity. In this paper, an improved face recognition system is developed to solve the problem of occlusion known as FRS-OCC. To build the FRS-OCC system, the color and texture features are used and then an incremental learning algorithm (Learn++) to select more informative features. Afterward, the trained stack-based autoencoder (SAE) deep learning algorithm is used to recognize a human face. Overall, the FRS-OCC system is used to introduce such algorithms which enhance the response time to guarantee a benchmark quality of service in any situation. To test and evaluate the performance of the proposed FRS-OCC system, the AR face dataset is utilized. On average, the FRS-OCC system is outperformed and achieved SE of 98.82%, SP of 98.49%, AC of 98.76% and AUC of 0.9995 compared to other state-of-the-art methods. The obtained results indicate that the FRS-OCC system can be used in any surveillance application.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권12호
• /
• pp.4439-4455
• /
• 2021

Compared with two-stage object detection algorithms, one-stage algorithms provide a better trade-off between real-time performance and accuracy. However, these methods treat the intermediate features equally, which lacks the flexibility to emphasize meaningful information for classification and location. Besides, they ignore the interaction of contextual information from different scales, which is important for medium and small objects detection. To tackle these problems, we propose an image pyramid network based on dual attention mechanism (DAIPNet), which builds an image pyramid to enrich the spatial information while emphasizing multi-scale informative features based on dual attention mechanisms for one-stage object detection. Our framework utilizes a pre-trained backbone as standard detection network, where the designed image pyramid network (IPN) is used as auxiliary network to provide complementary information. Here, the dual attention mechanism is composed of the adaptive feature fusion module (AFFM) and the progressive attention fusion module (PAFM). AFFM is designed to automatically pay attention to the feature maps with different importance from the backbone and auxiliary network, while PAFM is utilized to adaptively learn the channel attentive information in the context transfer process. Furthermore, in the IPN, we build an image pyramid to extract scale-wise features from downsampled images of different scales, where the features are further fused at different states to enrich scale-wise information and learn more comprehensive feature representations. Experimental results are shown on MS COCO dataset. Our proposed detector with a 300 × 300 input achieves superior performance of 32.6% mAP on the MS COCO test-dev compared with state-of-the-art methods.

• 복식문화연구
• /
• 제20권6호
• /
• pp.883-894
• /
• 2012

Fashion companies are faced with more severe competition with the emergence of new types of retail formats. Retailers are coming up with new shopping values to maximize their profits and benefits of customers. The aim of this study was to study shopping values and analyze differences in store selection criteria and store visits among. The respondents were males and females with ages ranging from the 20's to the 40's, residing in Seoul and the Gyeonggi area. Data were collected via both online and offline. Data from 427 respondents were analyzed using SPSS 17.0. Results indicated that there were three categories including hedonic, informative, and reliable shopping values from the factors for clothing shopping values. They form three types of consumer groups such as active, passive-reliable, and hedonic-informative shopping value groups. These three groups were different in terms of demographic characteristics. For the factor influencing store preference, the range of product selection and customer service were the two significant features that showed substantial differences in the shopping value groups store's atmosphere, salespeople, convenient location, price, and brand store did not have significant differences across groups. Retailers of each fashion retail formats have to consider consumers shopping values for their retail decision makings.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2003년도 ICCAS
• /
• pp.2731-2736
• /
• 2003

In highly informative, perception-rich environments that we call Omniscient Spaces, robots interact with physical objects which in turn afford robots the information showing how the objects should be manipulated. Object manipulation is commonly believed one of the most basic tasks in robot applications. However, no approaches including visual servoing seem satisfactory in unstructured environments such as our everyday life. Thus, in Omniscient Spaces, the features of the environments embed themselves in every entity, allowing robots to easily identify and manipulate unknown objects. To achieve this end, we propose a new paradigm of the interaction through Radio Frequency Identification (RFID). The aim of this paper is to learn about RFID and investigate how it works in object manipulation. Specifically, as an innovative trial for autonomous, real-time manipulation, a likely mobile robot equipped with an RFID system is developed. Details on the experiments are described together with some preliminary results.

• 한국해양공학회지
• /
• 제10권3호
• /
• pp.3-13
• /
• 1996

An extensive research programme has been aimed at the effct of ship weight distribution on the ship responses applying ship hydroelasticity theory. In the previous works, consistent tendencies of the still-water and the wave bending moments. respectively, were found as the weight distribution was varied systematically. The paper is therefore concerned mainly with any correlation between still-water and wave bending moments with the variations of weight distribution. Although these bending moments share different features with each other, such a comparison of tendencies was plausible and informative. These and other matters for the future are discussed.

• 한국산학기술학회논문지
• /
• 제14권8호
• /
• pp.3971-3975
• /
• 2013

본 논문은 핸드 제스쳐에 의해 증강현실 내의 가상 객체 제어기술로, HOG기반의 핸드 제스쳐 인식을 제안하고 있다. 인식을 위한 특징점들은 HOG불럭들에 의하여 결정되며, 크기가 다른 여러 불럭들을 시험하여 가장 적절한 불럭구성을 결정하며, AdaBoostSVM기법을 사용하여 분류 목적에 가장 적절한 불럭들을 추출한다. 실험 결과 핸드 제스쳐 인식률은 94% 이었다.

• 대한원격탐사학회지
• /
• 제28권6호
• /
• pp.671-682
• /
• 2012

Strong backscattering features from high-resolution Synthetic Aperture Rader (SAR) image provide useful information to analyze earth surface characteristics such as man-made objects in urban areas. The SAR image has, however, some limitations on description of detail information in urban areas compared to optical images. In this paper, we propose a new classification method using a fused SAR and Electro-Optical (EO) image, which provides more informative classification result than that of a single-sensor SAR image classification. The experimental results showed that the proposed method achieved successful results in combination of the SAR image classification and EO image characteristics.