• The Journal of the Acoustical Society of Korea
• /
• v.39 no.5
• /
• pp.497-504
• /
• 2020

Constructing speaker embeddings in speaker verification is an important issue. In general, a self-attention mechanism has been applied for speaker embedding encoding. Previous studies focused on training the self-attention in a high-level layer, such as the last pooling layer. In this case, the effect of low-level layers is not well represented in the speaker embedding encoding. In this study, we propose Masked Cross Self-Attentive Encoding (MCSAE) using ResNet. It focuses on training the features of both high-level and low-level layers. Based on multi-layer aggregation, the output features of each residual layer are used for the MCSAE. In the MCSAE, the interdependence of each input features is trained by cross self-attention module. A random masking regularization module is also applied to prevent overfitting problem. The MCSAE enhances the weight of frames representing the speaker information. Then, the output features are concatenated and encoded in the speaker embedding. Therefore, a more informative speaker embedding is encoded by using the MCSAE. The experimental results showed an equal error rate of 2.63 % using the VoxCeleb1 evaluation dataset. It improved performance compared with the previous self-attentive encoding and state-of-the-art methods.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.31 no.7A
• /
• pp.649-658
• /
• 2006

Recently, Software Development was applied to new-approach methods as a various form : client-server system and web-programing, object-orient concept, distributed development with a network environments. On the other hand, it be concerned about the distributed development technology and increasing of object-oriented methodology. These technology is spread out the software quality and improve of software production, reduction of the software develop working. Futures, we considered about the distributed software development technique with a many workstation. In this paper, we discussed optimal release problem based on a stochastic differential equation model for the distributed Software development environments. In the past, the software reliability applied to quality a rough guess with a software development process and approach by the estimation of reliability for a test progress. But, in this paper, we decided to optimal release times two method: first, SRGM with an error counting model in fault detection phase by NHPP. Second, fault detection is change of continuous random variable by SDE(stochastic differential equation). Here, we decide to optimal release time as a minimum cost form the detected failure data and debugging fault data during the system test phase and operational phase. Especially, we discussed to limitation of reliability considering of total software cost probability distribution.

• Journal of Biomedical Engineering Research
• /
• v.25 no.6
• /
• pp.439-445
• /
• 2004

Diffusion tensor image(DTI) exploits the random diffusional motion of water molecules. This method is useful for the characterization of the architecture of tissues. In some tissues, such as muscle or cerebral white matter, cellular arrangement shows a strongly preferred direction of water diffusion, i.e., the diffusion is anisotropic. The degree of anisotropy is often represented using diffusion anisotropy indices (relative anisotropy(RA), fractional anisotropy(FA), volume ratio(VR)). In this study, FA images were obtained using different gradient schemes(N=6, 11, 23, 35. 47). Mean values and the standard deviations of FA were then measured at several anatomic locations for each scheme. The results showed that both mean values and the standard deviations of FA were decreased as the number of gradient directions were increased. Also, the standard error of ADC measurement decreased as the number of diffusion gradient directions increased. In conclusion, different gradient schemes showed a significantly different noise performance and the schem with more gradient directions clearly improved the quality of the FA images. But considering acquisition time of image and standard deviation of FA, 23 gradient directions is clinically optimal.

• Journal of Soil and Groundwater Environment
• /
• v.11 no.5
• /
• pp.1-8
• /
• 2006

The objective of this study is to assess impacts of sorption heterogeneity on the transport of leachate leaked from unlined landfill sites and is accomplished by examining the results from a series of Monte-Carlo simulations. For random distribution coefficient ($K_{d}$) fields with four different levels of heterogeneity ranging from homogeneous to highly heterogeneous, the transport of leachate was investigated by linking a saturated flow model with a contaminant transport model. Impacts of a chemical heterogeneity were evaluated using point statistics values such as mean, standard deviation, and coefficient of variation of the concentration obtained at monitoring wells from 100 Monte-Carlo trials. Inspection of point statistics shows that the distribution of distribution coefficient in the landfill site proves to be an important parameter in controlling leachate concentrations. In comparison to homogeneous sorption, heterogeneous $K_{d^-}$ fields produce the variability in the leachate concentration for different realizations. The variability increases significantly as the variance in the $K_{d^-}$ field and the travel time between source and monitoring well increase. These outcomes indicate that use of a constant homogeneous $K_{d}$ value for predicting the transport of leachate can result in significant error, especially when variability in $K_{d}$ is high.

• Journal of agriculture & life science
• /
• v.46 no.5
• /
• pp.57-63
• /
• 2012

The purpose of this study was to estimate genetic parameters for productive traits in Duroc breed. In this study, 40,657 records for productive traits and the pedigree data of 47,974 families were collected from 41 farms registered at the Korean Animal Improvement Association (KAIA) from 2004 to 2011. The REMLf90 program was used to analyze a multiple traits animal model with fixed effects of sex, contemporary group, parity and age at the end of the test as covariate and random effects of animal and residual error. The heritabilities of days to 90 kg (D90KG), average daily gain (ADG), backfat thickness (BF) and eye muscle areas (EMA) were estimated to be 0.334, 0.340, 0.335, and 0.200, respectively. The genetic correlation coefficients were -0.992 between D90KG and ADG, -0.142 between ADG and BF, -0.361 between ADG and EMA, and -0.243 between BF and EMA. Conversely, positive genetic correlations for D90KG with BF and EMA were 0.13 and 0.36, respectively.

• Korean Journal of Remote Sensing
• /
• v.37 no.6_1
• /
• pp.1719-1729
• /
• 2021

Leaf area index (LAI) provides valuable information necessary for sustainable and effective management of forests. Although global high resolution LAI data are provided by European Space Agency using Sentinel-2 satellite images, they have not considered forest characteristics in model development and have not been evaluated for various forest ecosystems in South Korea. In this study, we proposed a LAI estimation model combining machine learning and the PROSAIL radiative transfer model using Sentinel-2 satellite data over a local forest area in South Korea. LAI-2200C was used to measure in situ LAI data. The proposed LAI estimation model was compared to the existing Sentinel-2 LAI product. The results showed that the proposed model outperformed the existing Sentinel-2 LAI product, yielding a difference of bias ~ 0.97 and a difference of root-mean-square-error ~ 0.81 on average, respectively, which improved the underestimation of the existing product. The proposed LAI estimation model provided promising results, implying its use for effective LAI estimation over forests in South Korea.

• Asian-Australasian Journal of Animal Sciences
• /
• v.32 no.9
• /
• pp.1349-1354
• /
• 2019

Objective: The objective of this study was to determine the best approach for handling missing records of first to successful insemination (FS) in Japanese Black heifers. Methods: Of a total of 2,367 records of heifers born between 2003 and 2015 used, 206 (8.7%) of open heifers were missing. Four penalty methods based on the number of inseminations were set as follows: C1, FS average according to the number of inseminations; C2, constant number of days, 359; C3, maximum number of FS days to each insemination; and C4, average of FS at the last insemination and FS of C2. C5 was generated by adding a constant number (21 d) to the highest number of FS days in each contemporary group. The bootstrap method was used to compare among the 5 methods in terms of bias, mean squared error (MSE) and coefficient of correlation between estimated breeding value (EBV) of non-censored data and censored data. Three percentages (5%, 10%, and 15%) were investigated using the random censoring scheme. The univariate animal model was used to conduct genetic analysis. Results: Heritability of FS in non-censored data was $0.012{\pm}0.016$, slightly lower than the average estimate from the five penalty methods. C1, C2, and C3 showed lower standard errors of estimated heritability but demonstrated inconsistent results for different percentages of missing records. C4 showed moderate standard errors but more stable ones for all percentages of the missing records, whereas C5 showed the highest standard errors compared with noncensored data. The MSE in C4 heritability was $0.633{\times}10^{-4}$, $0.879{\times}10^{-4}$, $0.876{\times}10^{-4}$ and $0.866{\times}10^{-4}$ for 5%, 8.7%, 10%, and 15%, respectively, of the missing records. Thus, C4 showed the lowest and the most stable MSE of heritability; the coefficient of correlation for EBV was 0.88; 0.93 and 0.90 for heifer, sire and dam, respectively. Conclusion: C4 demonstrated the highest positive correlation with the non-censored data set and was consistent within different percentages of the missing records. We concluded that C4 was the best penalty method for missing records due to the stable value of estimated parameters and the highest coefficient of correlation.

• Asian-Australasian Journal of Animal Sciences
• /
• v.32 no.7
• /
• pp.1015-1026
• /
• 2019

Objective: The objective of this study was to evaluate three different degrees of white striping (WS) addressing their automatic assessment and customer acceptance. The WS classification was performed based on a computer vision system (CVS), exploring different machine learning (ML) algorithms and the most important image features. Moreover, it was verified by consumer acceptance and purchase intent. Methods: The samples for image analysis were classified by trained specialists, according to severity degrees regarding visual and firmness aspects. Samples were obtained with a digital camera, and 25 features were extracted from these images. ML algorithms were applied aiming to induce a model capable of classifying the samples into three severity degrees. In addition, two sensory analyses were performed: 75 samples properly grilled were used for the first sensory test, and 9 photos for the second. All tests were performed using a 10-cm hybrid hedonic scale (acceptance test) and a 5-point scale (purchase intention). Results: The information gain metric ranked 13 attributes. However, just one type of image feature was not enough to describe the phenomenon. The classification models support vector machine, fuzzy-W, and random forest showed the best results with similar general accuracy (86.4%). The worst performance was obtained by multilayer perceptron (70.9%) with the high error rate in normal (NORM) sample predictions. The sensory analysis of acceptance verified that WS myopathy negatively affects the texture of the broiler breast fillets when grilled and the appearance attribute of the raw samples, which influenced the purchase intention scores of raw samples. Conclusion: The proposed system has proved to be adequate (fast and accurate) for the classification of WS samples. The sensory analysis of acceptance showed that WS myopathy negatively affects the tenderness of the broiler breast fillets when grilled, while the appearance attribute of the raw samples eventually influenced purchase intentions.

• Journal of Music and Human Behavior
• /
• v.16 no.1
• /
• pp.73-88
• /
• 2019

This research was conducted to compare the rhythm reproduction abilities between older adults with and without mild cognitive impairment (MCI) and analyze the abilities depending on the rhythm idiom. Participants between 60-85 years of age were recruited from senior community centers, dementia prevention centers, and senior welfare centers. A total of 57 participants were included in this study: 27 diagnosed with MCI and 30 healthy older adults (HOA). The experiment was conducted individually in a private room in which a participant was given random binary time rhythm idioms and instructed to reproduce the rhythmic idioms with finger tapping. Each participant's beat production was recorded with the Beat Processing Device (BPD) for iPad. BPD calculated rhythm reproduction as measured through rhythm ratio and error among beats. Results showed marginal differences between the two groups in terms of mean scores of rhythm reproduction abilities. In terms of the rhythm ratio among beats, both groups' highest rhythm reproduction rate was for <♩ ♩>, and their lowest reproduction rate was for <♩. ♪>. In conclusion, there was no significant difference in rhythm reproduction ability between the HOA and MCI groups. However, the study found an interesting result related to performance level of rhythmic idioms. This result provides therapeutic insight for formulating rhythm tasks for older adults.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.9
• /
• pp.89-96
• /
• 2021

In this paper, we analyzed that the user authentication scheme for TMIS(Telecare Medicine Information System) proposed by Al-Saggaf et al. In 2019, Al-Saggaf et al. proposed authentication scheme using biometric information, Al-Saggaf et al. claimed that their authentication scheme provides high security against various attacks along with very low computational cost. However in this paper after analyzing Al-Saggaf et al's authentication scheme, the Al-Saggaf et al's one are missing random number s from the DB to calculate the identity of the user from the server, and there is a design error in the authentication scheme due to the lack of delivery method. Al-Saggaf et al also claimed that their authentication scheme were safe against a variety of attacks, but were vulnerable to password guessing attack using login request messages and smart cards, session key exposure and insider attack. An attacker could also use a password to decrypt the stored user's biometric information by encrypting the DB with a password. Exposure of biometric information is a very serious breach of the user's privacy, which could allow an attacker to succeed in the user impersonation. Furthermore, Al-Saggaf et al's authentication schemes are vulnerable to identity guessing attack, which, unlike what they claimed, do not provide significant user anonymity in TMIS.