• The Journal of Internal Korean Medicine
• /
• v.40 no.3
• /
• pp.506-516
• /
• 2019

Objectives: This study aimed to report five patients who had multiple rib fractures after a traffic accident who improved with the combination of Korean medical admission treatment. Methods: We collected the data of traffic accident patients with multiple rib fractures who were admitted to the Daejeon Jaseng Hospital of Korean Medicine from April 2018 to May 2018 to receive the combinational Korean medical treatment. We observed these cases of patients treated by Acupuncture, Pharmacopuncture, Herbal medicine, Oriental physical therapy, Chuna treatment. We measured the validity of the treatment with a numerical rating scale (NRS) and the European Quality of Life-5 Dimension (EQ5D) at admission, at two weeks, and at the discharge date of hospitalization. Results: At the end of the treatment, all patients showed a decrease in NRS scores and in increase in EQ5D. The median NRS score was 6 (5-7) at the date of admission and 4 (2-7) at two weeks and then decreased to 3 (2-6). The median EQ5D score was 0.513 (0.350-0.752) at the date of admission and 0.692 (0.418-0.913) at two weeks, and then increased to 0.783 (0.671-0.913). Conclusions: After the combination of Korean medicine admission treatment, five patients with multiple rib fractures after a traffic accident showed that the treatments were effective. However, the number of subjects was insufficient and individual efficacy was not measured in this study. Therefore, further studies are needed on this topic.

• Speech Sciences
• /
• v.12 no.3
• /
• pp.115-124
• /
• 2005

We investigated on the submission of multiple candidates in speaker-independent speech recognition by VQ/HMM. Submission of fixed number of multiple candidates has first been examined. As the number of candidates increases by two, three, and four, the recognition error rates were found to decrease by 41%, 58%, and 65%, respectively compared to that of a single candidate. We tried another approach that the candidates within a range of Viterbi scores are submitted. The number of candidates showed geometric increase as the admitted range becomes large. For a practical application, a combination of the above two methods was also studied. We chose the candidates within some range of Viterbi scores and limited the maximum number of candidates submitted to five. Experimental results showed that recognition error rates of less than 10% could be achieved with average number of candidates of 3.2 by this method.

• Communications for Statistical Applications and Methods
• /
• v.14 no.3
• /
• pp.531-540
• /
• 2007

When multiple objectives exist, there are three approaches exist. These are maximin design, compound design, and constrained design. Still, each of three design criteria has its own strength and weakness. In this paper Hybrid approach is suggested when multiple design objectives exist, which is a combination of maximin and constrained design. Sometimes experimenter has several objectives, but he/she has only one or two primary objectives, others less important. A new approach should be useful under this condition. The genetic algorithm is used for few examples. It has been proven to be a very useful technique for this complex situation. Conclusion follows.

• IE interfaces
• /
• v.18 no.3
• /
• pp.247-252
• /
• 2005

In the automated production environment in the present days, the minimization of manual operation becomes a very important factor in increasing the efficiency of the production system. The exit burr produced through the milling operation on the edge of workpiece usually requires manual deburring process to enhance the level of precision of the resulting product. So far, researchers have developed various methods to understand the formation of exit burr in cutting process. One method to analytically identify the formation of exit burr was to use the geometrical information of CAD and CAM data used in automated machining. This method, in turn, generated the information resulting from the analysis such as burr type, cutting region, and exit angle. Up to now, the geometrical data were restricted to the single feature and single path. In this paper, a method to deal with the complicated geometric features such as line segment, arc, hole, and spline will be presented and validated using the field data. This method also deals with the complex workpiece shape which is a combination of multiple features. As for the cutting path, multiple tool path is analyzed in order to simulate the real cutting process. All this analysis is combined into a Windows based software and real data are used to validate the program in the conclusion.

• Proceedings of the KSRS Conference
• /
• 2007.10a
• /
• pp.6-10
• /
• 2007

Classification of hyperspectral images is challenging. A very high dimensional input space requires an exponentially large amount of data to adequately and reliably represent the classes in that space. In other words in order to obtain statistically reliable classification results, the number of necessary training samples increases exponentially as the number of spectral bands increases. However, in many situations, acquisition of the large number of training samples for these high-dimensional datasets may not be so easy. This problem can be overcome by using multiple classifiers. In this paper we compared the effectiveness of two approaches for creating multiple classifiers, feature selection and feature extraction. The methods are based on generating multiple feature subsets by running feature selection or feature extraction algorithm several times, each time for discrimination of one of the classes from the rest. A maximum likelihood classifier is applied on each of the obtained feature subsets and finally a combination scheme was used to combine the outputs of individual classifiers. Experimental results show the effectiveness of feature extraction algorithm for generating multiple classifiers.

• Journal of Korea Multimedia Society
• /
• v.22 no.6
• /
• pp.655-664
• /
• 2019

In this paper, we propose a novel way of combining multiple deep convolutional neural network (DCNN) architectures which work well for accurate video face identification by adopting a serial combination of 3D and 2D DCNNs. The proposed method first divides an input video sequence (to be recognized) into a number of sub-video sequences. The resulting sub-video sequences are used as input to the 3D DCNN so as to obtain the class-confidence scores for a given input video sequence by considering both temporal and spatial face feature characteristics of input video sequence. The class-confidence scores obtained from corresponding sub-video sequences is combined by forming our proposed class-confidence matrix. The resulting class-confidence matrix is then used as an input for learning 2D DCNN learning which is serially linked to 3D DCNN. Finally, fine-tuned, serially combined DCNN framework is applied for recognizing the identity present in a given test video sequence. To verify the effectiveness of our proposed method, extensive and comparative experiments have been conducted to evaluate our method on COX face databases with their standard face identification protocols. Experimental results showed that our method can achieve better or comparable identification rate compared to other state-of-the-art video FR methods.

• Journal of KIISE:Software and Applications
• /
• v.29 no.3
• /
• pp.192-201
• /
• 2002

Automated text classification is considered as an important method to manage and process a huge amount of documents in digital forms that are widespread and continuously increasing. Recently, text classification has been addressed with machine learning technologies such as k-nearest neighbor, decision tree, support vector machine and neural networks. However, only few investigations in text classification are studied on real problems but on well-organized text corpus, and do not show their usefulness. This paper proposes and analyzes text classification methods for a real application, email document classification task. First, we propose a combining method of multiple neural networks that improves the performance through the combinations with maximum and neural networks. Second, we present another strategy of combining multiple machine learning classifiers. Voting, Borda count and neural networks improve the overall classification performance. Experimental results show the usefulness of the proposed methods for a real application domain, yielding more than 90% precision rates.

• The Research Journal of the Costume Culture
• /
• v.27 no.5
• /
• pp.524-537
• /
• 2019

The aim of this study was to explore the effect of combinations of diverse methods notifying price discounts (i.e., reference price, odd price, and discount rate signs) and the relationships among product attribute perception, discount perception, attitude toward product, and purchase intention of product. Experiments were conducted where 12 stimuli of different price discount information notifications regarding T-shirt advertisements were presented to 352 informants. The results showed that notification of each type of discount information increased discount perception, whereas no effect due to the size of letters used in the discount rate notification was found. As more price discount information notifications were used, discount perception tended to become stronger. The results of ANOVA analysis show that both product attribute perception and discount perception affected attitude toward the product. In addition, product purchase intention was determined by attitude toward the product as well as price discount perception. Based on these findings, marketers may want to use a combination of methods of price discount notifications in advertisements to deliver price discount information clearly to consumers. Confirmation of discount information using multiple cues would help consumers to notice and perceive price discount information provided by retailers more effectively. Discount information is crucial for increasing both purchase intention and favorable attitude, therefore, diverse strategies regarding discount information presentations should be developed, tested and applied in the real world of retailing.

• Journal of the Earthquake Engineering Society of Korea
• /
• v.24 no.3
• /
• pp.141-148
• /
• 2020

For the important safety system, two or more units of identical equipment or redundant components with similar function were installed to prevent abnormal failure. If the failure probability of such equipment is independent, this redundancy could increase the system safety remarkably. However, if the failure of each component is highly correlated by installing in a structure or experiencing an earthquake event, the expected redundancy effect will decrease. Therefore, the seismic correlation of the equipment should be evaluated quantitatively for the seismic probabilistic safety assessment. The correlation effect can be explained in the procedure of constructing fragility curves. In this study, several methodologies to quantify the seismic correlation in the failure probability calculation for multiple components were reviewed and two possible ways considering the realistic situation were selected. Simple examples were tested to check the applicability of these methods. The conversion method between these two methods was suggested to render the evaluation using the advantages of each method possible.

• Journal of Positioning, Navigation, and Timing
• /
• v.4 no.1
• /
• pp.33-41
• /
• 2015

Compact Network Real-Time Kinematic (RTK) is a method that combines compact RTK and network RTK, and it can effectively reduce the time and spatial de-correlation errors. A network RTK user receives multiple correction information generated from reference stations that constitute a network, calculates correction information that is appropriate for one's own position through a proper combination method, and uses the information for the estimation of the position. This combination method is classified depending on the method for modeling the GPS error elements included in correction information, and the user position accuracy is affected by the accuracy of this modeling. Among the GPS error elements included in correction information, tropospheric delay is generally eliminated using a tropospheric model, and a combination method is then applied. In the case of a tropospheric model, the estimation accuracy varies depending on the meteorological condition, and thus eliminating the tropospheric delay of correction information using a tropospheric model is limited to a certain extent. In this study, correction information modeling accuracy performances were compared focusing on the Low-Order Surface Model (LSM), which models the GPS error elements included in correction information using a low-order surface, and a modified LSM method that considers tropospheric delay characteristics depending on altitude. Both of the two methods model GPS error elements in relation to altitude, but the second method reflects the characteristics of actual tropospheric delay depending on altitude. In this study, the final residual errors of user measurements were compared and analyzed using the correction information generated by the various methods mentioned above. For the performance comparison and analysis, various GPS actual measurement data were collected. The results indicated that the modified LSM method that considers actual tropospheric characteristics showed improved performance in terms of user measurement residual error and position domain residual error.