• 전자공학회논문지B
• /
• 제30B권9호
• /
• pp.84-93
• /
• 1993

Recently, several nonlinear shape normalization methods have been proposed in order to compensate for the shape distortions in handwritten characters. In this paper, we review these nonlinear shape normalization methods from the two points of view : feature projection and feature density equalization. The former makes feature projection histogram by projecting a certain feature at each point of input image into horizontal-or vertical-axis and the latter equalizes the feature densities of input image by re-sampling the feature projection histogram. A systematic comparison of these methods has been made based on the following criteria: recognition rate, processing speed, computational complexity and measure of variation. Then, we present the result of quantitative evaluation of each method based on these criteria for a large variety of handwritten Hangul syllables.

• ETRI Journal
• /
• 제38권3호
• /
• pp.502-509
• /
• 2016

Feature detection and description are key ingredients of common image processing and computer vision applications. Most existing algorithms focus on robust feature matching under challenging conditions, such as inplane rotations and scale changes. Consequently, they usually fail when the scene is blurred by camera shake or an object's motion. To solve this problem, we propose a new feature description algorithm that is robust to image blur and significantly improves the feature matching performance. The proposed algorithm builds a feature descriptor by considering the integral projection along four angular directions ($0^{\circ}$, $45^{\circ}$, $90^{\circ}$, and $135^{\circ}$) and by combining four projection vectors into a single highdimensional vector. Intensive experiment shows that the proposed descriptor outperforms existing descriptors for different types of blur caused by linear motion, nonlinear motion, and defocus. Furthermore, the proposed descriptor is robust to intensity changes and image rotation.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2005년도 ICCAS
• /
• pp.842-847
• /
• 2005

This paper proposes a novel real-time EMG pattern recognition for the control of a multifunction myoelectric hand from four channel EMG signals. To cope with the nonstationary signal property of the EMG, features are extracted by wavelet packet transform. For dimensionality reduction and nonlinear mapping of the features, we also propose a linear-nonlinear feature projection composed of PCA and SOFM. The dimensionality reduction by PCA simplifies the structure of the classifier, and reduces processing time for the pattern recognition. The nonlinear mapping by SOFM transforms the PCA-reduced features to a new feature space with high class separability. Finally a multilayer neural network is employed as the pattern classifier. We implement a real-time control system for a multifunction virtual hand. From experimental results, we show that all processes, including virtual hand control, are completed within 125 msec, and the proposed method is applicable to real-time myoelectric hand control without an operation time delay.

• 전자공학회논문지SC
• /
• 제42권2호
• /
• pp.39-48
• /
• 2005

본 논문에서는 다기능 근전의수를 제어하기 위해 전완에서 취득한 4 채널의 근전도로부터 9 가지 동작을 인식하는 새로운 방법을 제안한다. 비정상 신호특성을 가진 근전도를 해석하기 위해서 시간-주파수 영역에서 표현되는 특징벡터를 웨이블렛 패킷변환을 통해 추출한다. 높은 차원을 가지는 시간-주파수 특징벡터에 대하여 차원축소와 비선형변환을 수행하기 위해 PCA와 SOFM으로 구성된 특징투영 방법을 제안한다. PCA를 이용한 차원축소는 패턴분류기의 구조를 단순화하고 패턴인식을 위한 계산시간을 단축할 수 있다. SOFM을 이용한 비선형변환은 PCA에 의해 차원이 축소된 특징벡터를 새로운 공간으로 투영함으로써 클래스 분리도를 향상시킨다. 마지막으로 각 동작은 패턴분류기인 다층 신경회로망에 의해 인식된다. 실험 결과로부터 제안한 방법이 높은 인식률을 보임과 동시에 연속적인 패턴인식을 위한 실시간 구현이 가능함을 보인다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제12권7호
• /
• pp.3194-3216
• /
• 2018

Slow Feature Discriminant Analysis (SFDA) is a supervised feature extraction method inspired by biological mechanism. In this paper, a novel method called Two Dimensional Slow Feature Discriminant Analysis via $L_{2,1}$ norm minimization ($2DSFDA-L_{2,1}$) is proposed. $2DSFDA-L_{2,1}$ integrates $L_{2,1}$ norm regularization and 2D statically uncorrelated constraint to extract discriminant feature. First, $L_{2,1}$ norm regularization can promote the projection matrix row-sparsity, which makes the feature selection and subspace learning simultaneously. Second, uncorrelated features of minimum redundancy are effective for classification. We define 2D statistically uncorrelated model that each row (or column) are independent. Third, we provide a feasible solution by transforming the proposed $L_{2,1}$ nonlinear model into a linear regression type. Additionally, $2DSFDA-L_{2,1}$ is extended to a bilateral projection version called $BSFDA-L_{2,1}$. The advantage of $BSFDA-L_{2,1}$ is that an image can be represented with much less coefficients. Experimental results on three face databases demonstrate that the proposed $2DSFDA-L_{2,1}/BSFDA-L_{2,1}$ can obtain competitive performance.

• 한국정보처리학회논문지
• /
• 제7권5호
• /
• pp.1361-1369
• /
• 2000

For the implementation of 3D Sound Localization system, the binaural filtering by HRTFs is generally employed. But the HRTF filter is of high order and its coefficients for all directions have to be stored, which imposes a rather large memory requirement. To cope with this, research works have centered on obtaining low dimensional HRTF representations without significant loss of information and synthesizing the original HRTF efficiently, by means of feature extraction methods for multivariate dat including PCA. In these researches, conventional linear PCA was applied to the frequency domain HRTF data and using relatively small number of principal components the original HRTFs could be synthesized in approximation. In this paper we applied neural network based nonlinear PCA model (NLPCA) and the nonlinear PLS repression model (NLPLS) for this low dimensional HRTF modeling and analyze the results in comparison with the PCA. The NLPCA that performs projection of data onto the nonlinear surfaces showed the capability of more efficient HRTF feature extraction than linear PCA and the NLPLS regression model that incorporates the direction information in feature extraction yielded more stable results in synthesizing general HRTFs not included in the model training.

• 한국정보통신학회:학술대회논문집
• /
• 한국해양정보통신학회 2006년도 춘계종합학술대회
• /
• pp.784-787
• /
• 2006

본 논문에서는 한글 인쇄체 문자의 자소를 비선형적으로 분리하는 방법을 제안한다. 자소 분리 대상 문자는 자소의 조합 방식에 따라 6개의 형식으로 분류한다. 인쇄체 한글의 6형식 분류를 위해 그레이 레벨의 문자 이미지로부터 망 특성과 수직 수평 투영 기법을 이용해 특징을 추출하고, 오류 역전파 기법을 이용하여 분류를 시도한다. 분류된 문자 형식을 기반으로 분리 후보 영역을 지정하고, 이 영역을 기반으로 다단식 그래프 탐색 알고리즘을 이용하여 최적의 비선형적 자소 분리 경로를 찾아낸다. 실험 결과, 제안한 방법은 한글의 6형식 분류에 적합하였으며, 자소가 서로 붙어 선형적으로 분리가 어려운 문자의 자소 분리에 좋은 성능을 나타내었다.

• 한국통신학회논문지
• /
• 제32권9C호
• /
• pp.861-866
• /
• 2007

본 논문은 가우시안 확률분포함수 (Gaussian Probability Distribution Function) 데이터 군집화를 위해 중심신경망 (Centroid Neural Network, CNN)에 Bhattacharyya 커널을 적용한 군집화 알고리즘 (Bhattacharyya Kernel based CNN, BK-CNN)을 제안한다. 제안된 BK-CNN은 무감독 알고리즘인 중심신경망을 기반으로 하고 있으며, 커널 방법을 이용하여 데이터를 특징공간에서 투영한다. 입력공간의 비선형 문제를 선형적으로 해결하기 위해 제안한 커널 방법인데, 확률분포 사이의 거리측정을 위해 Bhattacharyya 거리를 이용한 커널방법을 사용하였다. 제안된 BK-CNN을 영상데이터 분류의 문제에 적용했을 때, 제안된 BK-CNN 알고리즘이 Bhattacharyya 커널을 적용한 k-means, 자기조직지도(Self-Organizing Map)와 중심 신경망등의 기존 알고리즘보다 1.7% - 4.3%의 평균 분류정확도 향상을 가져옴을 확인할 수 있었다.

• 한국IT서비스학회지
• /
• 제7권2호
• /
• pp.149-164
• /
• 2008

As the computing landscape is shifting to ubiquitous computing environments, there is increasingly growing the demand for a variety of device controls that react to user's implicit activities without excessively drawing user attentions. We developed an EMG-based car interface that enables the physically handicapped to drive a car using their functioning peripheral nerves. Our method extracts electromyogram signals caused by wrist movements from four places in the user's forearm and then infers the user's intent from the signals using multi-layered neural nets. By doing so, it makes it possible for the user to control the operation of car equipments and thus to drive the car. It also allows the user to enter inputs into the embedded computer through a user interface like an instrument LCD panel. We validated the effectiveness of our method through experimental use in a car built with the EMG-based interface.