• The Journal of the Acoustical Society of Korea
• /
• v.24 no.7
• /
• pp.410-415
• /
• 2005

Recently, most of the speaker verification systems are based on the pattern recognition approach method. And performance of the pattern-classifier depends on how to classify a variety of speakers' feature parameters. In order to classify feature parameters efficiently and effectively, it is of great importance to enlarge variations between speakers and effectively measure distances between feature parameters. Therefore, this paper would suggest the positively mixed model scheme that can enlarge inter-speaker variation by searching the individual model with world model at the same time. During decision procedure, we can maximize inter-speaker variation by using the proposed mixed model scheme. We also make use of a symbol probability weighting function in this system so as to reduce vector quantization errors by measuring symbol probability derived from the distance rate of between the world codebook and individual codebook. As the result of our experiment using this method, we could halve the Detection Cost Function (DCF) of the system from $2.37\%\;to\;1.16\%$.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.2
• /
• pp.219-224
• /
• 2020

Binary classification has been broadly investigated in machine learning. In addition, binary classification can be easily extended to multi class problems. To successfully utilize machine learning methods for classification tasks, preprocessing and feature extraction steps are essential. These are important steps to improve their classification performances. In this paper, we propose a new learning method based on weighted least squares. In the weighted least squares, designing weights has a significant role. Due to this necessity, we also propose a new technique to obtain weights that can achieve feature transformation. Based on this weighting technique, we also propose a method to combine the learning and feature extraction processes together to perform both processes simultaneously in one step. The proposed method shows the promising performance on five UCI machine learning data sets.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.36 no.6
• /
• pp.611-620
• /
• 2018

With the revitalization of alley market area becoming a major goal of the urban regeneration project, an understanding on customer characteristics that affect the sales of alley market areas is needed. As spatial heterogeneity appears to exist in alley market areas, the use of GWR (Geographically Weighted Regression) is required as an alternative to OLS (Ordinary Least Squares) regression. This study analyzes effects of customer characteristics on sales of 1007 alley market areas in Seoul. Comparing R squared and AICc, results show that GWR is better than OLS regression. According to OLS regression, the ratio of female, the ratio of 40's and 50's, the number of employees, the opening rate of establishment, the density of building and the size of alley market area have positive effects on sales, while the ratio of 20's and 30's, the distance of bus stop and that of subway station have negative effects. As a result of comparing local regression coefficients of geographically weighted regression analysis, the ratio of female customers has the greatest effect on the northwestern region, followed by the southwestern region, the central region and the northeastern region. The ratio of 20's and 30's and that of 40's and 50's effect on the southeastern and northeastern regions, and then the southwestern region. It is expected that this study will help to identify marketing target for each alley market area.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.39 no.4
• /
• pp.369-374
• /
• 2015

In order to efficiently optimize the problem involving complex computer codes or computationally expensive simulation, surrogate models are widely used. Because their accuracy significantly depends on sample points, many experimental designs have been proposed. One approach is the sequential design of experiments that consider existing information of responses. In earlier research, the correlation coefficients of the kriging surrogate model are introduced as weight parameters to define the scaled distance between sample points. However, if existing information is incorrect or lacking, new sample points can be misleading. Thus, our goal in this paper is to propose a weight function derived from correlation coefficients to generate new points robustly. To verify the performance of the proposed method, several existing sequential design methods are compared for use as mathematical examples.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.27 no.2
• /
• pp.126-134
• /
• 1990

In this study, eight function discrimination and recognition of the EMG signal from the biceps and triceps of 4 subjects were executed, using the Euclidean and weighted cepstral distance measure with LPC cepstrum coefficients. In case of Euclidean cepstral distance measure, as the number of LPC cepstrum coefficients was increased in 8, 10, 12, 14 the recognition rates of functions are 94.69, 95.63, 96.56, and 96.88[%], respectively, but increasing rates of recognition were inclined to decrease. In case of weighted cepstral distance measure, when the number of LPC cepstrum coefficients was 8, 10, 12 and 14, the recognition rates of functions were 91.88, 95, 99.69, and 96.63[%], respectively.

• Journal of the Korea Society of Computer and Information
• /
• v.11 no.1 s.39
• /
• pp.273-282
• /
• 2006

In this paper, we propose index method introducing distance concept in word by a method weighting word. This index method is frequent representing an inquiry word and document index and compound noun or more than two adjoin nouns or noun phrase, the farther the distance between these nouns, the fewer selected ratio decreases in index point is the aiming, this choose guide word candidate by existent weight grant method and distance between candidates chose candidate finally in index within 3 sentences. Using in these way I document of 100 kinds of newspaper, scientific treatise, web document and so on, showed the correctness rate resulted of newspaper 92.03% scientific treatise 95% web document 73.33%.

• Bulletin of the Korean Space Science Society
• /
• 2011.04a
• /
• pp.22.2-22.2
• /
• 2011

이 연구에서는 한반도 상공의 전리층 총전자수를 격자 형태로 나타냈다. 이를 위해 국토해양부 GPS 상시관측소에서 제공 중인 코드와 위상 측정값을 선형조합하였으며 그 결과물을 이용하여 시선방향 총전자수를 산출하였다. 이때 전리층 총전자수 산출결과의 정확도를 향상시키기 위해 가중최소자승법을 이용하여 위성과 수신기의 하드웨어 오차인 DCB(Differencial Code Bias)를 추정하였으며 추정된 DCB값은 IGS에서 제공 중인 DCB값과 비교하여 정확도를 확인하였다. 산출된 시선방향 총전자수를 연직방향 총전자수로 변환하기 위해 사상함수를 적용하였으며, 이를 다시 각 격자점에서의 연직방향 총전자수로 변환하기 위해 기존 연직방향 총전자수에 역거리 가중 보간법을 적용하였다. 각 격자점에서의 총전자수는 IGS(International GNSS Service)에서 제공 중인 GIM(Global Ionosphere Map) 모델의 총전자수와 비교하여 정확도를 확인하였다. 산출된 총전자수는 2시간 간격으로 나타내어 한반도 상공 전리층 총전자수의 변화 경향을 확인하였다.

• Proceedings of the Korean Statistical Society Conference
• /
• 2003.05a
• /
• pp.149-150
• /
• 2003

소표본 분할표 자료에서 적합도 검정통계량들의 카이제곱 근사 적용 가능에 대하여 많은 연구가 진행되었다. 소표본에서 세 가지 검정 통계량(피어슨 카이제곱 $X^{2}$, 일반화 가능도비 $G^{2}$, 그리고 역발산 I(2/3) 검정통계량)에 관하여 비교한 Rudas(1986)의 연구를 확장하여, 최근에 제안된 차이측도(BWHD(1/9), BWCS(1/3), NED(4/3) 검정통계량)를 포함시켜 비교 분석하였다. 독립모형의 이차원 분할표, 조건부 독립모형과 한 변수 독립 모형을 따르는 삼차원 분할표에 대한 모의실험을 통하여 생성된 90과 95 백분위수와 이에 대응하는 95% 신뢰구간을 살펴보고 실제 백분위수와 비교하였다. 그 결과 $X^{2}$, I(2/3), 그리고 BWHD(1/9) 검정통계량이 유사한 결과를 나타내었고 이 통계량들이 기존에 제안된 검정통계량들보다 적은 표본크기에서도 카이제곱 근사방법에 적용 가능함을 발견하였다.

• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 2011.04a
• /
• pp.292-293
• /
• 2011

보로노이 다면체 기법은 금속비정질재료에서의 20면체 Short-Range order와 같은 원자 구조를 분석하는 가장 유용한 도구 중에 하나이다. 기존의 보로노이 다면체 기법에서는 다면체의 평면은 다른 크기를 가진 원자간 거리의 중간에 위차하였다. 하지만, 이러한 방법은 필연적으로 분석 오차를 발생시킨다. 대안적인 방법으로서 근평면법이라고 불리우는 가중 보로노이법을 사용하면, 원자의 크기의 차이를 고려하여 해석할 수 있다. 본논문에서는 이러한 근평면법을 여러 가지의 성분비를 가진 비정질 재료에 적용하여 기존의 보로노이법과 비교 분석하였다. 결과적으로 보로노이 인덱스, 원자체적 분포, 보로노이 다면체 분포, 20면체의 공간 분포등등에서 기존의 보로노이법은 완전히 잘못된 결과를 보여줄 수도 있다는 것이 발견되었다.

• The Journal of the Acoustical Society of Korea
• /
• v.21 no.6
• /
• pp.552-558
• /
• 2002

Recently, based on the human speech recognition (HSR) model of Fletcher, the multi-band speech recognition has been intensively studied by many researchers. As a new automatic speech recognition (ASR) technique, the multi-band speech recognition splits the frequency domain into several sub-bands and recognizes each sub-band independently. The likelihood scores of sub-bands are weighted according to reliabilities of sub-bands and re-combined to make a final decision. This approach is known to be robust under noisy environments. When the noise is stationary a sub-band SNR can be estimated using the noise information in non-speech interval. However, if the noise is non-stationary it is not feasible to obtain the sub-band SNR. This paper proposes the inverse sub-band distance (ISD) weighting, where a distance of each sub-band is calculated by a stochastic matching of input feature vectors and hidden Markov models. The inverse distance is used as a sub-band weight. Experiments on 1500∼1800㎐ band-limited white noise and classical guitar sound revealed that the proposed method could represent the sub-band reliability effectively and improve the performance under both stationary and non-stationary band-limited noise environments.