• 대한전자공학회논문지SP
• /
• 제38권1호
• /
• pp.97-103
• /
• 2001

현재 HMM은 음성인식에서 가장 널리 쓰이는 방법이다. 대부분의 경우 HMM의 매개변수는 훈련데이터에 대해 최대유사도를 가지도록 훈련된다. 그러나 이러한 방법은 다른 단어들에 대한 변별력을 고려하지 않는 단점이 있다. 이 논문에서는 이러한 단점을 보완하기 위해, 유사단어에 대한 정보와 두 단어 사이에 변별력을 가지는 함수를 사용하여, 인식된 단어와 유사단어만을 대상으로 재인식하는 과정을 통해 단어를 검증하는 방법을 제안하였다. 유사단어는 각 단어의 HMM에 다른 단어의 훈련음성으로 확률값을 계산하여 가장 유사한 단어를 얻었으며, 단어간에 변별력을 가지는 인식기는 각 상태에 하중값을 가지는 인식기를 사용하여 구현하였다. 단어간에 변별력을 가지는 하중값은 유전자 알고리듬을 사용하여 얻었다. 실험에서 유사단어와 변별력을 가지는 검증기의 사용으로 오인식률이 약 22% 감소하였다.

• Communications for Statistical Applications and Methods
• /
• 제4권1호
• /
• pp.229-241
• /
• 1997

We consider discrimination curve and minimum dwell time for Poisson distribution and Poisson-power function distribution. Let the random variable X has Poisson distribution with mean .lambda.. For the hypothesis testing H$\_$0/:.lambda. = t vs. H$\_$1/:.lambda. = d (d$\_$0/ if X.leq.c. Since a critical value c can not be determined to satisfy both types of errors .alpha. and .beta., we considered discrimination curve that gives the maximum d such that it can be discriminated from t for a given .alpha. and .beta.. We also considered an algorithm to compute the minimum dwell time which is needed to discriminate at the given .alpha. and .beta. for the Poisson counts and proved its convergence property. For the Poisson-power function distribution, we reject H$\_$0/ if X.leq..'{c}.. Since a critical value .'{c}. can not be determined to satisfy both .alpha. and .beta., similar to the Poisson case we considered discrimination curve and computation algorithm to find the minimum dwell time for the Poisson-power function distribution. We prosent this algorithm and an example of computation. It is found that the minimum dwell time algorithm fails for the Poisson-power function distribution if the aiming error variance .sigma.$\^$2/$\_$2/ is too large relative to the variance .sigma.$\^$2/$\_$1/ of the Gaussian distribution of intensity. In other words, if .ell. is too small, we can not find the minimum dwell time for a given .alpha. and .beta..

• Communications for Statistical Applications and Methods
• /
• 제14권3호
• /
• pp.541-549
• /
• 2007

Testing the sibling relationship becomes more important issue in many cases such as the reunion of dispersed family members whose parents have already passed away and the discrimination between pure-bred and crossbred dogs. Analysis of the sibling case is different from that of the paternity case. In this paper, we describe how to evaluate and determine the sibling relationship by comparing sibling pairs with unrelated pairs. We use the Korean population with 17 independent STR loci system to propose a discrimination rule.

• Journal of Electrical Engineering and Technology
• /
• 제7권3호
• /
• pp.281-287
• /
• 2012

The customer load profile clustering method is used to make the TDLP (Typical Daily Load Profile) to estimate the quarter hourly load profile of non-AMR (Automatic Meter Reading) customers. This study examines how the repeated clustering method improves the ability to discriminate among the TDLPs of each cluster. The k-means algorithm is a well-known clustering technology in data mining. Repeated clustering groups the cluster into sub-clusters with the k-means algorithm and chooses the sub-cluster that has the maximum average error and repeats clustering until the final cluster count is satisfied.

• 한국시뮬레이션학회논문지
• /
• 제14권3호
• /
• pp.147-154
• /
• 2005

In conventional approaches such as a functinal approach or a parametric approach to online signature verification, which could not deal with the local shape of signature, much various important informations inherent in the local part of signature shape have been overlooked. In this paper, we try a structural approach in which a signature is represented as a structural form of handwriting primitives and the local parts along a signature handwriting can be selectively compared according to their discrimination power in the process of signature verification, As a result, the error rate is diminished in the case that the weights of subpattern units is applied into comparing process, which is the degree of discrimination power of local part. And also, the global variation and complexity of each signature extracted from the analysis of local shape is found useful in determining the decision threshold more precisely.

• 응용통계연구
• /
• 제17권1호
• /
• pp.75-88
• /
• 2004

서포트 벡터 기계(Support Vector Machines, SVMs) 알고리즘은 표본 점들과 분리 초평면 사이의 최소 거리를 최대화하는 것에 관심을 가져왔다. 본 논문은 모든 데이터 점들과 분리 초평면 사이의 거리들을 고려하는 total margin을 제안한다. 본 논문에서 제안하는 방법은 기존의 서포트 벡터 기계 알고리즘을 확장하고, 일반화 오차 경계를 개선하게 된다. 새롭게 제안하는 total margin알고리즘이 기존 방법들과의 비교를 통하여 더욱 우수한 수행능력을 가지고 있음을 수치 예제들을 통하여 확인할 수 있다.

• The Journal of the Acoustical Society of Korea
• /
• 제13권1E호
• /
• pp.64-70
• /
• 1994

예측신경회로망 모델은 패턴 예측에 의한 매우 효과적인 음성인식 모델이다. 그러나, 그러한 모델은 유사한 어휘간에서 변별력이 떨어지는 단점이 있다. 이 논문에서는 그러한 단점을 극복하기 위한 변별력있는 학습 알고리즘을 제안한다. 이 알고리즘은 최소 분류 오차 수식화와 GPD 알고리즘으로부터 유도외면 그에 따라서 인식 오차의 수를 직접 최소화하는 것이 가능하다. 한국어 숫자음에 대한 인식 실험결과, 기존의 알고리즘에서 발생하는 오인식의 30%를 줄일 수 있었다.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2015년도 추계학술대회
• /
• pp.1045-1046
• /
• 2015

When using the Support Vector Machine in the online signature verification, SVM kernel function should be chosen to use non-linear SVM and the constant parameters in the kernel functions should be adjusted to appropriate values to reduce the error rate of signature verification. Non-linear SVM which is built on a strong mathematical basis shows better performance of classification with the higher discrimination power. However, choosing the kernel function and adjusting constant parameter values depend on the heuristics of the problem domain. In the signature verification, this paper deals with the problems of selecting the correct kernel function and constant parameters' values, and shows the kernel function and coefficient parameter's values with the minimum error rate. As a result of this research, we expect the average error rate to be less than 1%.

• 한국근적외분광분석학회:학술대회논문집
• /
• 한국근적외분광분석학회 2001년도 NIR-2001
• /
• pp.1529-1529
• /
• 2001

The watercore in apple is very important factor in storage and sorting of fruit. Most consumers tend to prefer the apple included watercore in immediately after harvest, however the watercore causes fruit flesh to brown during storage and lose the worth after all. But it is practically impossible to judge to the naked eye whether an apple has watercore or not. Therefore, the rapid, accurate and non-destructive analysis method for discrimination of watercore should be settled without delay. In this study we attempted the discrimination and quantitative analysis of watercore in apple fruit using near-infrared transmittance spectroscopy ‘Fuji’ apple fruits produced in Kyungpook of Korea was used in this experiment. The watercore content in apple was evaluated by graphic treatment of culled slice sections(10 mm). NIR transmittance spectra were collected over the 500 to 1000 nm spectral region with a spectrometer (Sentronic Co., Germany). The calibration models were carried out by partial least squares (PLS) analysis between NIR spectra data of apples and chemical data of watercore content. The spectra were different in absorbance between apple included watercore and not included one. Apple included watercore had higher absorption band than sample not included one at 732 and 820 nm. The calibration model seems to be accurate to predict the watercore content in apple fruit, the correlation coefficient (R) and root mean square error of prediction (RMSEP) were 0.99 and 0.93%, respectively. This result indicates that the PLSR calibration model by using NIR transmittance spectroscopy could be used for discrimination of watercore in apple fruit.

• 대한인간공학회지
• /
• 제13권1호
• /
• pp.37-46
• /
• 1994

The accurate error size and discrimination region in the perception of depth amount from 3-dimensional images by the human visual system will be the basic data for the utilization and application of the binocular 3- eimensional image system. This paper is focused on studying the accuracy of the depth amount perceived from 3- dimensional images by the human visual system. From the performed experiment, the following results have been obtained: (1) The depth amount perceived from the binocular 3- dimensional images has been displayed by a proper scale of distance, and found to be imprecise and also have a large variance. (2) In utilizing the binocular 3-dimensional image system, it seems more appropriate to make the images viewed outward rather than inward from the screen in the regard of error and variance. (3) The binocular 3-dimensional image system can be effectively applied to displaying unreal space, for example, the layout of room in design, from the viewpoint of perception characteristics of depth amount.