• The Journal of the Acoustical Society of Korea
• /
• v.29 no.1
• /
• pp.76-81
• /
• 2010

In this paper, we propose a novel approach to improve the performance of a statistical model-based voice activity detection (VAD) which is based on the second-order conditional maximum a posteriori (CMAP). In our approach, the VAD decision rule is expressed as the geometric mean of likelihood ratios (LRs) based on adapted threshold according to the speech presence probability conditioned on both the current observation and the speech activity decisions in the pervious two frames. Experimental results show that the proposed approach yields better results compared to the statistical model-based and the CMAP-based VAD using the LR test.

• The Korean Journal of Applied Statistics
• /
• v.24 no.3
• /
• pp.477-484
• /
• 2011

When an event of target variable is rare, a widespread strategy is to build a model on the sample that disproportionally over-represents the events, that is over-sampled. Using the data over-sampled from the original data set, the predicted values would be biased; however, it can be easily corrected to represent the population. In this study, we investigate into the relationship between the proportion of rare event on a data-mart and the model performance using real world data of a Korean credit card company. Also, we use the methods for adjusting of posterior probability for over-sampled data of the offset method and the weighted method. Finally, we compare the performance of the methods using real data sets.

• Magazine of the Korea Concrete Institute
• /
• v.11 no.2
• /
• pp.241-249
• /
• 1999

To estimate the compressive strength of concrete more realistically, relative large number of data are necessary. However, it is very common in practice that only limited data are available. The purpose of the present paper is therefore to propose a realistic method to estimate the compressive strength of concrete with limited data in actual site. The Bayesian method of statistical analysis has been applied to the problem of the estimation of compressive strength of concrete. The mean compressive strength is considered as the random parameter and a prior distribution is selected to enable updating of the Bayesian distribution of compressive strength of concrete reflecting both existing data and sampling observations. The updating of the Bayesian distribution with increasing data is illustrated in numerical application. It is shown that by combining prior estimation with information from site observation, more precise estimation is possible with relatively small sampling. It is also seen that the contribution of the prior in determining the posterior distribution depends on its sharpness or flatness in relation to the sharpness or flatness of the likelihood function. The present paper allows more realistic determination of concrete strength in site with limited data.

• Proceedings of the Acoustical Society of Korea Conference
• /
• 1998.06c
• /
• pp.499-502
• /
• 1998

본 논문에서는 한 쌍의 수동소나를 이용하여 미지의 잠항물체의 존재 유무를 확인하고 각 센서에 도달하는 시간지연의 차를 평가하는 Detection과 Tracking 알고리즘을 연구한다. 이 과정에서 이동하는 표적의 속력에 의한 도플러효과를 보상하는 2차원 확률분포 함수를 적용함으로 보다 정확한 결과를 도출한다. 관측신호의 Cross-Correlation과 Bayesian Method를 이용하여 계산한 시간지연과 도플러효과 비의 이차원 Likelihood 함수로부터 사후확률 (Posterior Probability)을 구하여 발견 평가와 추적을 수행한다.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.23 no.10
• /
• pp.1195-1200
• /
• 2019

This paper proposes a detection algorithm for group testing. Group testing is a problem of finding a very small number of defect samples out of a large number of samples, which is similar to the problem of Compressed Sensing. In this paper, we define a noiseless group testing and propose a probabilistic algorithm for detection of defective samples. The proposed algorithm is constructed such that the extrinsic probabilities between the input and output signals exchange with each other so that the posterior probability of the output signal is maximized. Then, defective samples are found in the group testing problem through a simulation on the detection algorithm. The simulation results for this study are compared with the lower bound in the information theory to see how much difference in failure probability over the input and output signal sizes.

• Proceedings of the Korean Statistical Society Conference
• /
• 2002.11a
• /
• pp.169-174
• /
• 2002

본 연구는 k개 지수분포 모수들의 기하평균에 대한 베이지안추정 방법을 제시하였다. 이를 위해 Tibshirani가 제안한 직교변환법으로 비정보적 사전확률분포를 도출하여 모수들의 결합사후확률분포를 유도해 내었으며, 이 분포 하에서 가중 몬테칼로 방법을 사용하여 기하평균을 추정하는 절차를 제안하였다. 모의실험과 실제자료의 예를 통해 제안된 베이지안 추정의 유효성 및 효용성을 보였으며, 본 연구에서 제안한 사전확률분포가 전통적인 포함확률을 기준으로 볼 때, Jeffrey의 사전확률분포 보다 더 유효한 추정을 함을 보였다.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.34 no.6C
• /
• pp.588-592
• /
• 2009

In this paper, we propose a novel method to improve the performance of the global soft decision which is based on the second-order conditional maximum a posteriori (CMAP). Conventional global soft decision scheme has an disadvantage in that the speech absence probability adjusted by a fixed-parameter was sensitive to the various noise environments. In proposed approach using the second-order CMAP, speech absence probability value is more flexible which exploit not only the current observation but also the speech activity decisions in the previous two frames. Experimental results show that the proposed improved global soft decision method based on second-order conditional MAP yields better results compared to the conventional global soft decision technique with the performance criteria of the ITU-T P. 862 perceptual evaluation of speech quality (PESQ).

• The Korean Journal of Applied Statistics
• /
• v.14 no.2
• /
• pp.379-391
• /
• 2001

본 연구에서는 관심거리가 되고 있는 마코프인쇄 몬테칼로(Markov Chain Monte Carlo, MCMC)방법에 근거한 공간 확률난수 (spatial random variate)생성법과 깁스표본추출법(Gibbs sampling)에 의한 베이지안 분석 방법에 대한 기술적 사항들에 관하여 검토하였다. 먼저 기본적인 확률난수 생성법과 관련된 사항을 살펴보고, 다음으로 조건부명시법(conditional specification)을 이용한 공간 확률난수 생성법을 예를 들어 살펴보기로한다. 다음으로는 이렇게 생성된 공간자료를 분석하기 위하여 깁스표본추출법을 이용한 베이지안 사후분포를 구하는 방법을 살펴보았다.

• The KIPS Transactions:PartB
• /
• v.8B no.3
• /
• pp.231-236
• /
• 2001

일반적인 다층 신경망에서 가중치의 갱신 알고리즘으로 사용하는 오류 역전과 방식은 가중치 갱신 결과를 고정된(fixed) 한 개의 값으로 결정한다. 이는 여러 갱신의 가능성을 오직 한 개의 값으로 고정하기 때문에 다양한 가능성들을 모두 수용하지 못하는 면이 있다. 하지만 모든 가능성을 확률적 분포로 표현하는 갱신 알고리즘을 도입하면 이런 문제는 해결된다. 이러한 알고리즘을 사용한 베이지안 신경망 모형(Bayesian Neural Networks Models)은 주어진 입력값(Input)에 대해 블랙 박스(Black-Box)와같은 신경망 구조의 각 층(Layer)을 거친 출력값(Out put)을 계산한다. 이 때 주어진 입력 데이터에 대한 결과의 예측값은 사후분포(posterior distribution)의 기댓값(mean)에 의해 계산할 수 있다. 주어진 사전분포(prior distribution)와 학습데이터에 의한 우도함수(likelihood functions)에 의해 계산한 사후확률의 함수는 매우 복잡한 구조를 가짐으로 기댓값의 적분계산에 대한 어려움이 발생한다. 따라서 수치해석적인 방법보다는 확률적 추정에 의한 근사 방법인 몬테 칼로 시뮬레이션을 이용할 수 있다. 이러한 방법으로서 Hybrid Monte Carlo 알고리즘은 좋은 결과를 제공하여준다(Neal 1996). 본 논문에서는 Hybrid Monte Carlo 알고리즘을 적용한 신경망이 기존의 CHAID, CART 그리고 QUEST와 같은 여러 가지 분류 알고리즘에 비해서 우수한 결과를 제공하는 것을 나타내고 있다.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.46 no.4
• /
• pp.132-138
• /
• 2009

In this paper, we propose a novel approach to improve the performance of minima controlled recursive averaging (MCRA) which is based on the second-order conditional maximum a posteriori (CMAP). From an investigation of the MCRA scheme, it is discovered that the MCRA method cannot take full consideration of the inter-frame correlation of voice activity since the noise power estimate is adjusted by the speech presence probability depending on an observation of the current frame. To avoid this phenomenon, the proposed MCRA approach incorporates the second-order CMAP criterion in which the noise power estimate is obtained using the speech presence probability conditioned on both the current observation and the speech activity decisions in the previous two frames. Experimental results show that the proposed MCRA technique based on second-order conditional MAP yields better results compared to the conventional MCRA method.