• The Korean Journal of Applied Statistics
• /
• v.31 no.5
• /
• pp.539-554
• /
• 2018

Effective interaction between user and device is considered an important ability of IoT devices. For some applications, it is necessary to recognize human facial expressions in real time and make accurate judgments in order to respond to situations correctly. Therefore, many researches on facial image analysis have been preceded in order to construct a more accurate and faster recognition system. In this study, we constructed an automatic recognition system for facial expressions through two steps - a facial recognition step and a classification step. We compared various models with different sets of data with pixel information, landmark coordinates, Euclidean distances among landmark points, and arctangent angles. We found a fast and efficient prediction model with only 30 principal components of face landmark information. We applied several prediction models, that included linear discriminant analysis (LDA), random forests, support vector machine (SVM), and bagging; consequently, an SVM model gives the best result. The LDA model gives the second best prediction accuracy but it can fit and predict data faster than SVM and other methods. Finally, we compared our method to Microsoft Azure Emotion API and Convolution Neural Network (CNN). Our method gives a very competitive result.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2009.11a
• /
• pp.179-181
• /
• 2009

In this paper, we propose an adaptive noise detection and modified Gaussian removal algorithm using local statistics for impulse noise. In order to determine constraints for noise detection, the local mean, variance, and maximum values are used. In addition, a modified Gaussian filter that integrates the tuning parameter to remove the detected noises. Experimental results show that our method is significantly better than a number of existing techniques in terms of image restoration and noise detection.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.44 no.4 s.316
• /
• pp.78-85
• /
• 2007

This paper presents a segmental scheme for regions-composed images in local normalization process. The scheme is based on local statistics computed through a moving window. The normalization algorithm uses linear or nonlinear functions to transfer the pixel distribution and the homogeneous affine of regions which is corrupted by additive noise. It adjusts the mean and standard deviation for nearest-neighbor interpoint distance between current and the normalized image signals and changes the segmentation performance according to local statistics and parameter variation adaptively. The performance of newly advanced local normalization algorithm is evaluated and compared to the performance of conventional normalization methods. Experimental results are presented to show the region segmentation properties of these approaches.

• Journal of Computing Science and Engineering
• /
• v.8 no.2
• /
• pp.94-106
• /
• 2014

Schema matching is widely used in many applications, such as data integration, ontology merging, data warehouse and dataspaces. In this paper, we propose a novel matching technique that is based on the order of attributes appearing in the schema structure of query results. The appearance order embodies the extent of the importance of an attribute for the user examining the query results. The core idea of our approach is to collect statistics about the appearance order of attributes from the query logs, to find correspondences between attributes in the schemas to be matched. As a first step, we employ a matrix to structure the statistics around the appearance order of attributes. Then, two scoring functions are considered to measure the similarity of the collected statistics. Finally, a traditional algorithm is employed to find the mapping with the highest score. Furthermore, our approach can be seen as a complementary member to the family of the existing matchers, and can also be combined with them to obtain more accurate results. We validate our approach with an experimental study, the results of which demonstrate that our approach is effective, and has good performance.

• Communications for Statistical Applications and Methods
• /
• v.30 no.5
• /
• pp.501-516
• /
• 2023

Knowledge distillation (KD) refers to extracting knowledge from a large and complex model (teacher) and transferring it to a relatively small model (student). This can be done by training the teacher model to obtain the activation function values of the hidden or the output layers and then retraining the student model using the same training data with the obtained values. Recently, relational KD (RKD) has been proposed to extract knowledge about relative differences in training data. This method improved the performance of the student model compared to conventional KDs. In this paper, we propose a new method for RKD by introducing a new loss function for RKD. The proposed loss function is defined using the area difference between the teacher model and the student model in a specific hidden layer, and it is shown that the model can be successfully compressed, and the generalization performance of the model can be improved. We demonstrate that the accuracy of the model applying the method proposed in the study of model compression of audio data is up to 1.8% higher than that of the existing method. For the study of model generalization, we demonstrate that the model has up to 0.5% better performance in accuracy when introducing the RKD method to self-KD using image data.

• Journal of radiological science and technology
• /
• v.40 no.2
• /
• pp.197-203
• /
• 2017

The purpose of this study was to measure signal to noise ratio (SNR) according to change of equalization from region of interest (ROI) of heart shadow in chest image. We examined images of chest image of 87 patients in a University-affiliated hospital, Seoul, Korea. Chest images of each patient were calculated by using ImageJ. We have analysis socio-demographical variables, SNR according to images, 95% confidence according to SNR of difference in a mean of SNR. Differences of SNR among change of equalization were tested by SPSS Statistics21 ANOVA test for there was statistical significance 95%(p < 0.05). In SNR results, with the quality of distributions in the order of original chest image, original chest image heart shadow and equalization chest image, equalization chest image heart shadow(p < 0.001). In conclusion, this study would be that quantitative evaluation of heart shadow on chest image can be used as an adjunct to the histogram equalization chest image.

• Journal of the Korean Society of Radiology
• /
• v.11 no.6
• /
• pp.443-451
• /
• 2017

The purpose of this study was to comparison of measure signal to noise ratio (SNR) according to change of kernel size from region of interest (ROI) of heart shadow in chest image. We examined images of chest image of 100 patients in a University-affiliated hospital, Seoul, Korea. Chest images of each patient were calculated by using ImageJ. We have analysis socio-demographical variables, SNR according to images, 95% confidence according to SNR of difference in a mean of SNR. Differences of SNR among change of equalization were tested by SPSS Statistics21 ANOVA test for there was statistical significance 95%(p<0.05). In SNR results, with the quality of distributions in the order of kernel size 9*9 image, kernel size 7*7 image and original chest image, kernel size 3*3 image (p<0.001). In conclusion, this study would be that quantitative evaluation of heart shadow on chest image can be used as an adjunct to the kernel size chest image.

• Journal of Digital Convergence
• /
• v.16 no.7
• /
• pp.317-324
• /
• 2018

The purpose of this study was to investigate factors affecting body image in adolescents. The subject of study is 140 adolescents in the three middle schools from the J city. The data were collected using a self-report questionnaire from March 15 to March 29, 2017. Data were analyzed by descriptive statistics, T-test, ANOVA, Pearson's correlation, and multiple regression with SPSS 20.0 program. Body image was found to be in a significant positive correlation with family strength and self-esteem. The result of the multiple regression indicates the self-esteem(${\beta}=.24$, p<.05), family strength(${\beta}=.19$, p<.05) and gender(${\beta}=.16$, p<.05) predict 17.9% (F=11.10, p<.001) of body image. Therefore, it is necessary to develop self-esteem and family strength strategy intervention program for establishing a right body image of adolescents.

• Journal of the Korean Data and Information Science Society
• /
• v.19 no.3
• /
• pp.723-734
• /
• 2008

The main reason for deinterlacing is frame-rate conversion. The other reason for deinterlacing is of course improve clarity and reduce flicker. Using a deinterlacer can help clarity and stability of the image. Many deinterlacing algorithms are available in image processing literatures such as ELA and E-ELA. This paper propose a new statistical deinterlacing algorithm based on statistical tests such as the Bartlett test, the Levene test and the Kruskal-Wallis test. The results obtained from the proposed algorithms are found to be comparable to those from many well-known deinterlacers. However, the results in the proposed deinterlacers are found to be more efficient than other deinterlacers.

• Journal of KIISE:Software and Applications
• /
• v.32 no.8
• /
• pp.742-751
• /
• 2005

This paper proposes a novel text image watermarking algorithm based on word classification. The words are classified into K classes using simple features. Several adjacent words are grouped into a segment. and the segments are also classified using the word class information. The same amount of information is inserted into each of the segment classes. The signal is encoded by modifying some inter-word spaces statistics of segment classes. Subjective comparisons with conventional word-shift algorithms are presented under several criteria.