• Geomechanics and Engineering
• /
• v.34 no.1
• /
• pp.29-41
• /
• 2023

The 3D geospatial modeling of geotechnical information can aid in understanding the geotechnical characteristic values of the continuous subsurface at construction sites. In this study, a geostatistical optimization model for the three-dimensional (3D) mapping of subsurface stratification and the SPT-N value based on a trial-and-error rule was developed and applied to a dam emergency spillway site in South Korea. Geospatial database development for a geotechnical investigation, reconstitution of the target grid volume, and detection of outliers in the borehole dataset were implemented prior to the 3D modeling. For the site-specific subsurface stratification of the engineering geo-layer, we developed an integration method for the borehole and geophysical survey datasets based on the geostatistical optimization procedure of ordinary kriging and sequential Gaussian simulation (SGS) by comparing their cross-validation-based prediction residuals. We also developed an optimization technique based on SGS for estimating the 3D geometry of the SPT-N value. This method involves quantitatively testing the reliability of SGS and selecting the realizations with a high estimation accuracy. Boring tests were performed for validation, and the proposed method yielded more accurate prediction results and reproduced the spatial distribution of geotechnical information more effectively than the conventional geostatistical approach.

• The Journal of the Acoustical Society of Korea
• /
• v.38 no.6
• /
• pp.703-709
• /
• 2019

This paper proposes speech recognition systems employing Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM) structures combined with Hidden Markov Moldel (HMM) to effectively recognize the speech of VeloPharyngeal Insufficiency (VPI) patients, and compares the recognition performance of the systems to the Gaussian Mixture Model (GMM-HMM) and fully-connected Deep Neural Network (DNNHMM) based speech recognition systems. In this paper, the initial model is trained using normal speakers' speech and simulated VPI speech is used for generating a prior model for speaker adaptation. For VPI speaker adaptation, selected layers are trained in the CNN-HMM based model, and dropout regulatory technique is applied in the LSTM-HMM based model, showing 3.68 % improvement in recognition accuracy. The experimental results demonstrate that the proposed LSTM-HMM-based speech recognition system is effective for VPI speech with small-sized speech data, compared to conventional GMM-HMM and fully-connected DNN-HMM system.

• Nuclear Engineering and Technology
• /
• v.52 no.7
• /
• pp.1495-1502
• /
• 2020

A new method of X-ray source spectrum estimation based on compressed sensing is proposed in this paper. The algorithm K-SVD is applied for sparse representation. Nonnegative constraints are added by modifying the L₁ reconstruction algorithm proposed by Rosset and Zhu. The estimation method is demonstrated on simulated spectra typical of mammography and CT. X-ray spectra are simulated with the Monte Carlo code Geant4. The proposed method is successfully applied to highly ill conditioned and under determined estimation problems with a good performance of suppressing noises. Results with acceptable accuracies (MSE < 5%) can be obtained with 10% Gaussian white noises added to the simulated experimental data. The biggest difference between the proposed method and the existing methods is that multiple prior knowledge of X-ray spectra can be included in one dictionary, which is meaningful for obtaining the true X-ray spectrum from the measurements.

• The Journal of the Acoustical Society of Korea
• /
• v.35 no.2
• /
• pp.134-142
• /
• 2016

In this paper, a method synthesizing speech signal using the 40 kHz ultrasonic signals reflected from the articulatory muscles was introduced and performance was evaluated. When the ultrasound signals are radiated to articulating face, the Doppler effects caused by movements of lips, jaw, and chin observed. The signals that have different frequencies from that of the transmitted signals are found in the received signals. These ADS (Acoustic-Doppler Signals) were used for estimating of the speech parameters in this study. Prior to synthesizing speech signal, a quantitative correlation analysis between ADS and speech signals was carried out on each frequency bin. According to the results, the feasibility of the ADS-based speech synthesis was validated. ADS-to-speech transformation was achieved by the joint Gaussian mixture model-based conversion rules. The experimental results from the 5 subjects showed that filter bank energy and LPC (Linear Predictive Coefficient) cepstrum coefficients are the optimal features for ADS, and speech, respectively. In the subjective evaluation where synthesized speech signals were obtained using the excitation sources extracted from original speech signals, it was confirmed that the ADS-to-speech conversion method yielded 72.2 % average recognition rates.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2016.05a
• /
• pp.110-110
• /
• 2016

도시 유역의 강우-유출 모의에는 지표 투수율 및 하수관거 영향 등 인위적 배수계통의 영향을 고려할 수 있는 도시유출모형이 널리 이용되고 있으며, 모형 검증을 통해 모의 성능을 평가한다. 도시유출모형의 검증은 일반적인 강우-유출 모형과 같이 강우사상별 유량의 관측시계열과 모의시계열의 목적함수가 최소가 되는 최적 매개변수를 탐색하는 과정이다. 도시유출모형의 검증에서 발생하는 문제점은 크게 다음과 같다. 첫째, 대규모 도시 유역의 복잡하고 다양한 하수관거에 대한 최적매개변수를 관거별로 구하는 것은 물리적으로 불가능하다. 따라서 동일 배수분구내 하수관거의 매개변수 값은 동일하다고 가정하거나, 모형 단순화 과정을 통해 매개변수의 물리적 범위 내에서 최적해를 탐색해야 하는 단순화에서 기인한 불확실성이 있다. 둘째, 다양한 매개변수들의 물리적 범위를 고려하기 위해서는 전역최적화기법이 유효하다. 그러나 전역최적화 종류, 목적함수, 모의횟수, 목표성능별 최적 매개변수 결과가 각각 다르므로 추정된 최적 매개변수의 범위에 대한 불확실성이 있다. 이에 본 연구에서는 Bayesian 모형과 EPA SWMM(Storm Water Management Model)을 연계하여 도시유출모형의 매개변수 불확실성을 정량적으로 분석할 수 있는 모형을 제안하고자 한다. 이를 위해 서울 우이천 유역을 대상으로 SWMM 모형을 구축하고, 절단 정규분포(truncated Gaussian distribution)를 사전분포(prior)로 가정하여 매개변수의 물리적 범위를 고려하였다. 최종적으로 결합확률분포로 계산된 각 매개변수간 사후분포를 통해 모의된 유출량의 불확실성을 정량적으로 분석하였다. 본 연구에서 제안된 모형은 대규모 도시 유역의 도시유출모형 구축 시 다양한 매개변수의 물리적 범위를 고려한 최적화와 동시에 내재된 불확실성을 정량적으로 분석할 수 있으므로, 침수예측 및 홍수예경보 등의 문제에서 상당한 신뢰성을 확보할 수 있을 것으로 판단된다.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.6
• /
• pp.209-220
• /
• 2014

Compressive Sensing (CS) is a new signal acquisition paradigm which enables sampling under Nyquist rate for a special kind of signal called sparse signal. There are plenty of CS recovery methods but their performance are still challenging, especially at a low sub-rate. For CS recovery of natural images, regularizations exploiting some prior information can be used in order to enhance CS performance. In this context, this paper addresses improving quality of reconstructed natural images based on Dantzig selector and smooth filters (i.e., Gaussian filter and nonlocal means filter) to generate a new regularization called smooth residual error regularization. Moreover, total variation has been proved for its success in preserving edge objects and boundary of reconstructed images. Therefore, effectiveness of the proposed regularization is verified by experimenting it using augmented Lagrangian total variation minimization. This framework is considered as a new CS recovery seeking smoothness in residual images. Experimental results demonstrate significant improvement of the proposed framework over some other CS recoveries both in subjective and objective qualities. In the best case, our algorithm gains up to 9.14 dB compared with the CS recovery using Bayesian framework.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.48 no.4
• /
• pp.21-29
• /
• 2011

At existing skin detection methods using skin color information defined based on the prior knowldege, threshold value to be used at the stage of dividing the backround and the skin region was decided on a subjective point of view through experiments. Also, threshold value was selected in a passive manner according to their background and illumination environments in these existing methods. These existing methods displayed a drawback in that their performance was fully influenced by the threshold value estimated through repetitive experiments. To overcome the drawback of existing methods, this paper propose a skin region detection method using a histogram approximation based on the mean shift algorithm. The proposed method is to divide the background region and the skin region by using the mean shift method at the histogram of the skin-map of the input image generated by the comparison of the similarity with the standard skin color at the CbCr color space and actively finding the maximum value converged by brightness level. Since the histogram has a form of discontinuous function accumulated according to the brightness value of the pixel, it gets approximated as a Gaussian Mixture Model (GMM) using the Bezier Curve method. Thus, the proposed method detects the skin region by using the mean shift method and actively finding the maximum value which eventually becomes the dividing point, not by using the manually selected threshold value unlike other existing methods. This method detects the skin region high performance effectively through experiments.