• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1998.10a
• /
• pp.255-262
• /
• 1998

This paper presents a translation, rotation and scale invariant methodology for the recognition of closed planar shape images using the bispectrum of a contour sequence and the learning vector quantization(LVQ) neural classifier. The contour sequences obtained from the closed planar images represent the Euclidean distance between the centroid and all boundary pixels of the shape, and are related to the overall shape of the images. The higher order spectra based on third order cumulants is applied to tihs contour sample to extract fifteen bispectral feature vectors for each planar image. There feature vector, which are invariant to shape translation, rotation and scale transformation, can be used to represent two0dimensional planar images and are fed into a neural network classifier. The LVQ architecture is chosen as a neural classifier because the network is easy and fast to train, the structure is relatively simple. The experimental recognition processes with eight different hapes of aircraft images are presented to illustrate the high performance of this proposed method even the target images are significantly corrupted by noise.

• Annual Conference on Human and Language Technology
• /
• 2022.10a
• /
• pp.115-121
• /
• 2022

질문에 답하기 위해 관련 구절을 검색하는 기술은 오픈 도메인 질의응답의 검색 단계를 위해 필요하다. 전통적인 방법은 정보 검색 기법인 빈도-역문서 빈도(TF-IDF) 기반으로 희소한 벡터 표현을 활용하여 구절을 검색한다. 하지만 희소 벡터 표현은 벡터 길이가 길 뿐만 아니라, 질문에 나오지 않는 단어나 토큰을 검색하지 못한다는 취약점을 가진다. 밀집 벡터 표현 연구는 이러한 취약점을 개선하고 있으며 대부분의 연구가 영어 데이터셋을 학습한 것이다. 따라서, 본 연구는 한국어 데이터셋을 학습한 밀집 벡터 표현을 연구하고 여러 가지 부정 샘플(negative sample) 추출 방법을 도입하여 전이 학습한 모델 성능을 비교 분석한다. 또한, 대화 응답 선택 태스크에서 밀집 검색에 활용한 순위 재지정 상호작용 레이어를 추가한 실험을 진행하고 비교 분석한다. 밀집 벡터 표현 모델을 학습하는 것이 도전적인 과제인만큼 향후에도 다양한 시도가 필요할 것으로 보인다.

• Journal of KIISE:Computing Practices and Letters
• /
• v.14 no.1
• /
• pp.1-12
• /
• 2008

In this paper, we present a real-time algorithm for recognizing the vehicle color from the indoor and outdoor vehicle images based on GPU (Graphics Processing Unit) acceleration. In the preprocessing step, we construct feature victors from the sample vehicle images with different colors. Then, we combine the feature vectors for each color and store them as a reference texture that would be used in the GPU. Given an input vehicle image, the CPU constructs its feature Hector, and then the GPU compares it with the sample feature vectors in the reference texture. The similarities between the input feature vector and the sample feature vectors for each color are measured, and then the result is transferred to the CPU to recognize the vehicle color. The output colors are categorized into seven colors that include three achromatic colors: black, silver, and white and four chromatic colors: red, yellow, blue, and green. We construct feature vectors by using the histograms which consist of hue-saturation pairs and hue-intensity pairs. The weight factor is given to the saturation values. Our algorithm shows 94.67% of successful color recognition rate, by using a large number of sample images captured in various environments, by generating feature vectors that distinguish different colors, and by utilizing an appropriate likelihood function. We also accelerate the speed of color recognition by utilizing the parallel computation functionality in the GPU. In the experiments, we constructed a reference texture from 7,168 sample images, where 1,024 images were used for each color. The average time for generating a feature vector is 0.509ms for the $150{\times}113$ resolution image. After the feature vector is constructed, the execution time for GPU-based color recognition is 2.316ms in average, and this is 5.47 times faster than the case when the algorithm is executed in the CPU. Our experiments were limited to the vehicle images only, but our algorithm can be extended to the input images of the general objects.

• Journal of Korean Society for Quality Management
• /
• v.21 no.2
• /
• pp.215-223
• /
• 1993

A paired comparison which is a method of comparative judgements is widely used to increase the amount of information transmitted to human sense organ. In this study, a paired comparison method is proposed for the discrimination capacity and preference analysis of non-skilled sensory inspectors. We consider on order effect, that is, difference of evaluation occurred as a result of sample presentation by random order. The purpose of this paper is to determine a sample with the highest preference degrees by analyzing a discrimination capacity and evaluating preference degrees of sensory inspectors. An analysis of a discrimination capacity is bared on the capacity index obtained by an eigen-vector method and an evaluation of preference degrees is performed by a significance test.

• ETRI Journal
• /
• v.35 no.6
• /
• pp.990-1000
• /
• 2013

This paper describes the scalable extension of High Efficiency Video Coding (HEVC) to provide flexible high-quality digital video content services. The proposed scalable codec is designed on multi-loop decoding architecture to support inter-layer sample prediction and inter-layer motion parameter prediction. Inter-layer sample prediction is enabled by inserting the reconstructed picture of the reference layer (RL) into the decoded picture buffer of the enhancement layer (EL). To reduce the motion parameter redundancies between layers, the motion parameter of the RL is used as one of the candidates in merge mode and motion vector prediction in the EL. The proposed scalable extension can support scalabilities with minimum changes to the HEVC and provide average Bj${\o}$ntegaard delta bitrate gains of about 24% for spatial scalability and of about 21% for SNR scalability compared to simulcast coding with HEVC.

• Journal of Industrial Technology
• /
• v.31 no.B
• /
• pp.3-7
• /
• 2011

This paper addresses a computationally simple adaptive beamforming method to cancel interferences arriving onto a sensor array. In the proposed method, an estimate of the interference subspace is extracted from a submatrix of the sample covariance matrix and an orthonormal basis for the estimated subspace is efficiently found, one basis vector being updated every sample. Its computational burden is just $O(M{\eta})$ in an M-sensor array when ${\eta}$ directional signals are present. The new method does not make any premises of the geometrical structure of arrays, and can be applied to arbitrary arrays.

• Journal of the Korean Statistical Society
• /
• v.24 no.2
• /
• pp.303-312
• /
• 1995

We consider the following type of general semi-parametric non-linear regression model : $y_i = f_i(\theta) + \epsilon_i, i=1, \cdots, n$ where ${f_i(\cdot)}$ represents the set of non-linear functions of the unknown parameter vector $\theta' = (\theta_1, \cdots, \theta_p)$ and ${\epsilon_i}$ represents the set of measurement errors with unknown distribution. Under suitable finite-sample, small-dispersion asymptotic framework, we derive a general lower bound for the asymptotic mean squared error (AMSE) matrix of the Gauss-consistent estimator of $\theta$. We then prove the fundamental result that the general non-linear least squares estimator (NLSE) is an optimal estimator within the class of all regular Gauss-consistent estimators irrespective of the type of the distribution of the measurement errors.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.1 no.3
• /
• pp.4-19
• /
• 1991

In this paper, we propose a fuzzy neural network pattern classifier utilizing fuzzy information. This system works without any a priori information about the number of clusters or cluster centers. It classifies each input according to the distance between the weights and the normalized input using Bezdek's [1] fuzzy membership value equation. This model returns the correct membership value for each input vector and find several cluster centers. Some experimental studies of comparison with other algorithms will be presented for sample data sets.

• Journal of KIISE:Software and Applications
• /
• v.35 no.3
• /
• pp.162-168
• /
• 2008

A look-up table based vector adaptive filter is proposed in color inverse halftoning. Inverse halftoning converts halftone image into a continuous-tone image. The templates and training images are required in the process of look-up table based methods, which can be obtained from distributed patterns in the sample halftone images and their original images. Although the look-up table based methods usually are faster and show better performances in PSNR than other methods do, they show wide range of qualities depending on how they treat nonexisting patterns in the look-up table. In this paper, a vector adaptive filter is proposed to compensate for these nonexisting patterns, which achieves better quality owing to the contributed informations about hue, saturation, and intensity of surrounding pixels. The experimental results showed that the proposed method resulted in higher PSNR than that of conventional Best Linear Estimation method. The bigger the size of the template in the look-up table becomes, the more outstanding quality in the proposed method can be obtained.

• The Korean Journal of Applied Statistics
• /
• v.29 no.1
• /
• pp.61-73
• /
• 2016

We consider the wild bootstrap Ljung-Box (LB) test for autocorrelation in residuals of fitted multivariate time series models. The asymptotic chi-square distribution under the IID assumption is traditionally used for the LB test; however, size distortion tends to occur in the usage of the LB test, due to the conditional heteroskedasticity of financial time series. In order to overcome such defects, we propose the wild bootstrap LB test for autocorrelation in residuals of fitted vector autoregressive and error correction models. The simulation study and real data analysis are conducted for finite sample performance.