• The Journal of the Acoustical Society of Korea
• /
• v.28 no.1
• /
• pp.77-83
• /
• 2009

This paper proposes a new noise-biased compensation of minimum statistics(MS) method using a nonlinear function and a priori speech absence probability(SAP) for speech enhancement in non-stationary noisy environments. The minimum statistics(MS) method is well known technique for noise power estimation in non-stationary noisy environments. It tends to bias the noise estimate below that of true noise level. The proposed method is combined with an adaptive parameter based on a sigmoid function and a priori speech absence probability (SAP) for biased compensation. Specifically. we apply the adaptive parameter according to the a posteriori SNR. In addition, when the a priori SAP equals unity, the adaptive biased compensation factor separately increases ${\delta}_{max}$ each frequency bin, and vice versa. We evaluate the estimation of noise power capability in highly non-stationary and various noise environments, the improvement in the segmental signal-to-noise ratio (SNR), and the Itakura-Saito Distortion Measure (ISDM) integrated into a spectral subtraction (SS). The results shows that our proposed method is superior to the conventional MS approach.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.17 no.4
• /
• pp.965-970
• /
• 2013

It is important work to extract saliency regions from an image as preprocessing for various image processing methods. In this paper, we introduce an improved method to estimate saliency value of each pixel from an image. The proposed method is an improved work of the previously studied method using color and statistical framework to estimate saliency values. At first, saliency value of each pixel is calculated using the local contrast of an image region at various scales and the most significant saliency pixel is determined using saliency value of each pixel. Then, saliency value of each pixel is again estimated using gauss weight with respect to the most significant saliency pixel and the saliency of each pixel is determined to calculate initial probability. At last, the saliency value of each pixel is calculated by Bayes' rule. The experiments show that our approach outperforms the current statistical based method.

• The Journal of the Acoustical Society of Korea
• /
• v.17 no.4
• /
• pp.76-81
• /
• 1998

문장의 에너지 컨투어 제어를 위한 방법을 제안한다. 제안된 모델은 음절을 기본단 위로 하며 다층 신경망을 이용하여 어절내 각 음절의 피크값을 추정한다. 그리고 신경망 입 력에 문장내 어절의 기능과 관련된 간접적인 피라미터를 수용함으로써 문장단위의 에너지 컨투어 생성을 가능하게 하였다. 본 추정기의 예측오차는 학습문장에 대해 10% 이내로서 높은 신뢰도를 보인다. 또한 이용함으로써 단어사전 없이 단어의 에너지 컨투어 추정을 가 능하게 하였으며, TTS에의 활용 가능성을 보였다.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.40 no.10
• /
• pp.895-900
• /
• 2016

Remaining useful life (RUL) estimation of the Li-ion battery has gained great interest because it is necessary for quality assurance, operation planning, and determination of the exchange period. This paper presents the RUL estimation of an Li-ion battery for an energy storage system using exponential function for the degradation model and Markov Chain Monte Carlo (MCMC) approach for parameter estimation. The MCMC approach is dependent upon information such as model initial parameters and input setting parameters which highly affect the estimation result. To overcome this difficulty, this paper offers a guideline for model initial parameters based on the regression result, and MCMC input parameters derived by comparisons with a thorough search of theoretical results.

• Journal of Intelligence and Information Systems
• /
• v.22 no.3
• /
• pp.1-21
• /
• 2016

IT-based subjective well-being (SWB) services, a main part of wellness IT, should measure the SWB state of individuals in an unrestrained, cost-effective manner. The dictionaries for sentiment analysis available in the market may be useful for this purpose, but obtaining proper sentiment values using only words from the sentiment lexicon is impossible; therefore, a new dictionary including wellness vocabulary is needed. The existing sentiment dictionaries link only a single sentiment value to a single sentiment word, although sentiment values may vary depending on personal traits. In this study, we develop an extended version of the SenticNet sentiment dictionary dubbed WellnessWordNet. SenticNet is considered the best and most expressive among the already existing sentiment dictionaries. Using the information provided by SenticNet, we created a database including the wellness states (estimated values) of stress, depression, and anger to develop the WellnessWordNet system. The accuracy of the system was validated through actual tests with live subjects. This study is unique and unprecedented in that i) an extended sentiment dictionary, WellnessWordNet, is developed; ii) values for wellness state language are offered; and iii) different sentiment values, namely contextual polarity, for people of the same gender or age group are suggested.

• Journal of Korean Society of Transportation
• /
• v.23 no.7 s.85
• /
• pp.125-136
• /
• 2005

The goal of this paper is to develop freeway Origin-Destination (OD) demand estimation model using VDS and AVI data. The formulation of methodology proposed in this paper includes traffic flow technique to be able to remove the bi-level problem and optimal solution algorithm using a kalman filter algorithm. The proposed dynamic OD estimation model use ilk and off-ramp volumes collected from VDS and partial OD collected from AVI data to raise the accuracy of dynamic OD estimation. The proposed model is evaluated by using the real-time data of SOHAEAN freeway, South Korea. The result of the proposed dynamic OD estimation model based on VDS and AVI data is better than that of based on VDS data. The more AVI systems are equipped at on and off-ramp, the more excellent result of estimation accuracy is expected.

• The Korean Journal of Applied Statistics
• /
• v.33 no.4
• /
• pp.365-377
• /
• 2020

We discuss latent factor regression when constructing a common structure inherent among explanatory variables to solve multicollinearity and use them as regressors to construct a linear model of a response variable. Bayesian estimation with LASSO prior of a large penalty parameter to construct a significant factor loading matrix of intrinsic interests among infinite latent structures. The estimated factor loading matrix with estimated other parameters can be inversely transformed into linear parameters of each explanatory variable and used as prediction models for new observations. We apply the proposed method to Product Service Management data of HBAT and observe that the proposed method constructs the same factors of general common factor analysis for the fixed number of factors. The calculated MSE of predicted values of Bayesian latent factor regression model is also smaller than the common factor regression model.

• Proceedings of the Korean Statistical Society Conference
• /
• 2000.11a
• /
• pp.135-140
• /
• 2000

베이지안 신경망 모형(Bayesian Neural Networks Models)에서 주어진 입력값(input)은 블랙 박스(Black-Box)와 같은 신경망 구조의 각 층(layer)을 거쳐서 출력값(output)으로 계산된다. 새로운 입력 데이터에 대한 예측값은 사후분포(posterior distribution)의 기대값(mean)에 의해 계산된다. 주어진 사전분포(prior distribution)와 학습데이터에 의한 가능도함수(likelihood functions)를 통해 계산되어진 사후분포는 매우 복잡한 구조를 갖게 됨으로서 기대값의 적분계산에 대한 어려움이 발생한다. 이때 확률적 추정에 의한 근사 방법인 몬테칼로 적분을 이용한다. 이러한 방법으로서 Hybrid Monte Carlo 알고리즘은 우수한 결과를 제공하여준다(Neal 1996). 본 논문에서는 Hybrid Monte Carlo 알고리즘과 기존에 많이 사용되고 있는 Gibbs sampling, Metropolis algorithm, 그리고 Slice Sampling등의 몬테칼로 방법들을 비교한다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.07a
• /
• pp.667-670
• /
• 2020

최근에는 증강현실, 로봇공학 등의 분야에서 객체의 위치 검출 이외에도, 객체의 자세에 대한 추정이 요구되고 있다. 객체의 자세 정보가 포함된 데이터셋은 위치 정보만 포함된 데이터셋에 비하여 상대적으로 매우 적기 때문에 인공 신경망 구조를 활용하기 어려운 측면이 있으나, 최근에 들어서는 기계학습 기반의 자세 추정 알고리즘들이 여럿 등장하고 있다. 본 논문에서는 이 가운데 Dense 6d Pose Object detector (DPOD) [11]의 구조를 기반으로 하여 가구의 조립 설명서에 그려진 가구 부품들의 자세를 추정하고자 한다. DPOD [11]는 입력으로 RGB 영상을 받으며, 해당 영상에서 자세를 추정하고자 하는 객체의 영역에 해당하는 픽셀들을 추정하고, 객체의 영역에 해당되는 각 픽셀에서 해당 객체의 3D 모델의 UV map 값을 추정한다. 이렇게 픽셀 개수만큼의 2D - 3D 대응이 생성된 이후에는, RANSAC과 PnP 알고리즘을 통해 RGB 영상에서의 객체와 객체의 3D 모델 간의 변환 관계 행렬이 구해지게 된다. 본 논문에서는 사전에 정해진 24개의 자세 후보들을 기반으로 가구 부품의 3D 모델을 2D에 투영한 RGB 영상들로 인공 신경망을 학습하였으며, 평가 시에는 실제 조립 설명서에서의 가구 부품의 자세를 추정하였다. 실험 결과 IKEA의 Stefan 의자 조립 설명서에 대하여 100%의 ADD score를 얻었으며, 추정 자세가 자세 후보군 중 정답 자세에 가장 근접한 경우를 정답으로 평가했을 때 100%의 정답률을 얻었다. 제안하는 신경망을 사용하였을 때, 가구 조립 설명서에서 가구 부품의 위치를 찾는 객체 검출기(object detection network)와, 각 개체의 종류를 구분하는 객체 리트리벌 네트워크(retrieval network)를 함께 사용하여 최종적으로 가구 부품의 자세를 추정할 수 있다.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.47 no.1
• /
• pp.112-119
• /
• 2010

This paper defines the probability models for determining the disparity map given stereo images and derives the methods for solving the problem, which is proven to be equivalent to an energy-based stereo matching. Under the assumptions the difference between the pixel on the left image and the corresponding pixel on the right image and the difference between the disparities of the neighboring pixels are exponentially distributed, a recursive approach for estimating the MRF regularizing parameter is proposed. Usually energy-based stereo matching methods are so sensitive to the parameter that it should be carefully determined. The proposed method alternates between estimating the parameter with the intermediate disparity map and estimating the disparity map with the estimated parameter, after computing it with random initial parameter. It is shown that the parameter estimated by the proposed method converges to the optimum and its performance can be improved significantly by adjusting the parameter and modifying the energy term.