• Journal of Korean Society for Quality Management
• /
• v.22 no.3
• /
• pp.34-53
• /
• 1994

Taguchi's robust design methodology has focused only on a single performance characteristic or response, but the quality of most products is seldom defined by a characteristic, and is rather the composite of a family of characteristics which are often interrelated and nearly always measured in a variety of units. The multiple performance characteristics problem is how to compromise the conflicts among the selected levels of the design parameters for each individual performance characteristic. In this paper, the modified desirability function using SN ratio which can be optimized by univariate technique is proposed and a parameter design procedure to achieve the best balance among several different response variables is developed We reanalyze two existing case studies by the proposed method and compare these results with ones by the sum of SN ratios and the expected weighted loss.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.6
• /
• pp.1692-1705
• /
• 2024

In this paper, we present a method that integrates a Grammar Transducer as an external language model to enhance the accuracy of the pre-trained Korean End-to-end (E2E) Automatic Speech Recognition (ASR) model. The E2E ASR model utilizes the Connectionist Temporal Classification (CTC) loss function to derive hypothesis sentences from input audio. However, this method reveals a limitation inherent in the CTC approach, as it fails to capture language information from transcript data directly. To overcome this limitation, we propose a fusion approach that combines a clause-level n-gram language model, transformed into a Weighted Finite-State Transducer (WFST), with the E2E ASR model. This approach enhances the model's accuracy and allows for domain adaptation using just additional text data, avoiding the need for further intensive training of the extensive pre-trained ASR model. This is particularly advantageous for Korean, characterized as a low-resource language, which confronts a significant challenge due to limited resources of speech data and available ASR models. Initially, we validate the efficacy of training the n-gram model at the clause-level by contrasting its inference accuracy with that of the E2E ASR model when merged with language models trained on smaller lexical units. We then demonstrate that our approach achieves enhanced domain adaptation accuracy compared to Shallow Fusion, a previously devised method for merging an external language model with an E2E ASR model without necessitating additional training.

• Physical Therapy Korea
• /
• v.22 no.3
• /
• pp.98-105
• /
• 2015

Muscular dystrophy is a hereditary musculoskeletal disorder caused by a mutation in the dystrophin gene. Duchenne muscular dystrophy (DMD) is one of the most common, and progresses relatively faster than other muscular dystrophies. It is characterized by progressive myofiber degeneration, muscle weakness and ultimately ambulatory loss. Since it is an X-linked recessive inheritance, DMD is mostly expressed in males and rarely expressed or less severe in females. The most effective measurement tool for DMD is magnetic resonance imaging (MRI), which allows non-invasive examination of longitudinal measurement. It can detect progressive decline of skeletal muscle size by measuring a maximal cross-sectional area of skeletal muscle. Additionally, other techniques in MRI, like $T_2$-weighted imaging, assess muscle damage, including inflammation, by detecting changes in $T_2$ relaxation time. Current MRI techniques even allow quantification of metabolic differences between affected and non-affected muscles in DMD. There is no current cure, but physical therapist can improve their quality of life by maintaining muscle strength and function, especially if treatment (and other forms of medical intervention) begins in the early stages of the disease.

• Journal of Korean Society for Quality Management
• /
• v.24 no.3
• /
• pp.111-132
• /
• 1996

Taguchi's parameter design is to determine optimal settings of design parameters of a product or a process such that the characteristics of a product exhibit small variabilities around their target values. His analysis of the problem has focused only on a single characteristic or response, but the quality of most products is seldom defined by a characteristic, and is rather the composite of a great number of characteristics which are often interrelated and nearly always measured in a variety of units. The critical problem in dealing with multiple characteristics is how to compromise the conflicts among the selected levels of the design parameters for each individual characteristic. In this paper, the EXTOPSIS Model using SN ratio which can be optimized by univariate technique is proposed and a parameter design procedure to achieve the optimal compromise among several different response variables is developed. Two existing case studies are solved by the proposed method and the results are compared with ones by the sum of SN ratios, the expected weighted loss, and the desirability function.

• ETRI Journal
• /
• v.27 no.6
• /
• pp.777-787
• /
• 2005

In this paper, we propose an urgency- and efficiency-based wireless packet scheduling (UEPS) algorithm that is able to schedule real-time (RT) and non-real-time (NRT) traffics at the same time while supporting multiple users simultaneously at any given scheduling time instant. The UEPS algorithm is designed to support wireless downlink packet scheduling in an orthogonal frequency division multiple access (OFDMA) system, which is a strong candidate as a wireless access method for the next generation of wireless communications. The UEPS algorithm uses the time-utility function as a scheduling urgency factor and the relative status of the current channel to the average channel status as an efficiency indicator of radio resource usage. The design goal of the UEPS algorithm is to maximize throughput of NRT traffics while satisfying quality-of-service (QoS) requirements of RT traffics. The simulation study shows that the UEPS algorithm is able to give better throughput performance than existing wireless packet scheduling algorithms such as proportional fair (PF) and modified-largest weighted delay first (M-LWDF), while satisfying the QoS requirements of RT traffics such as average delay and packet loss rate under various traffic loads.

• Proceedings of the Korean Information Science Society Conference
• /
• 2005.07b
• /
• pp.226-228
• /
• 2005

Predicting the cellular location of an unknown protein gives valuable information for inferring the possible function of the protein. For more accurate Prediction system, we need a good feature extraction method that transforms the raw sequence data into the numerical feature vector, minimizing information loss. In this paper we propose new methods of extracting underlying features only from the sequence data by computing pairwise sequence alignment scores. In addition, we use composition based features to improve prediction accuracy. To construct an SVM ensemble from separately trained SVM classifiers, we propose specificity based weighted majority voting . The overall prediction accuracy evaluated by the 5-fold cross-validation reached $88.53\%$ for the eukaryotic animal data set. By comparing the prediction accuracy of various feature extraction methods, we could get the biological insight on the location of targeting information. Our numerical experiments confirm that our new feature extraction methods are very useful forpredicting subcellular localization of proteins.

• Journal of the Korean Society of Environmental Restoration Technology
• /
• v.19 no.1
• /
• pp.121-133
• /
• 2016

It was required to evaluate ecological restorations in a comprehensive way in order to systematically manage conservation areas such as DMZ and national parks in South Korea. In this research we developed a new approach to evaluating ecological restorations with more various indexes than vegetation covering-related indexes. By analyzing damaged areas in the vicinity of DMZ, major damaged types were identified as six classes: landform modification, surface loss, soil pollution, soil physio-chemical modification, vegetation decline and vegetation damaged. From literature review, 39 indexes were selected and were grouped into four divisions: soil property, vegetation growth & structure, habitat property and landscape structure & functions. By conducting a survey with the selected indexes targeting relevant experts, data on relative importance among the divisions and indexes by damaged type were collected. As a result, it was found that the orders and values of weighted values of the divisions were different by damaged type: for example, soil property (0.402), vegetation growth & structure (0.209), habitat property (0.225), landscape structure & function (0.163) for "landform modification"; but soil property (0.171), vegetation growth & structure (0.401), habitat property (0.270), landscape structure & function (0.158) for "vegetation decline". Similarly, evaluation indexes showed different orders and values of relative importance, easiness in field measurement and representativeness for the division by damaged type, and the values were used for calculating importance factor for each index. In the evaluation table, score1 and score2 were made by the importance factors of indexes multiplied by distribution values which present grades and by the weighted values of divisions. In conclusion, while dealing with the damaged type was considered significant for evaluating and managing restorations, further tests on this table with a range of cases were needed to improve its quality.

• IE interfaces
• /
• v.17 no.spc
• /
• pp.111-121
• /
• 2004

In this paper, we propose an urgency and efficiency based wireless packet scheduling (UEPS) algorithm that is able to schedule real time (RT) and non-real time (NRT) traffics at the same time. The proposed UEPS algorithm is designed to support wireless downlink packet scheduling in the OFDMA system which is a strong candidate wireless system for the next generation mobile communications. The UEPS algorithm uses the time-utility function as a scheduling urgency factor and the relative status of the current channel to the average one as an efficiency indicator of radio resource usage. The design goal of the UEPS algorithm is to maximize throughput of NRT traffics with satisfying QoS requirements of RT traffics. The simulation study shows that the proposed UEPS algorithm is able to give better throughput performance than existing wireless packet scheduling algorithms such as proportional fair (PF) and modified-largest weighted delay first (M-LWDF) while satisfying QoS requirements of RT traffics such as the average delay and the packet loss rate under various traffic loads.

• The Journal of the Acoustical Society of Korea
• /
• v.19 no.2
• /
• pp.33-39
• /
• 2000

We have studied to obtain the frequency characteristics of the Surface Acoustic Wave(SAW) bandpass filter, having low shape factor, it's interdigital transducer(IDT) was formed on the 35° Y-cut X-propagation Quartz substrate and was evaporated by Aluminium. And then, we performed computer-simulation by a simulator. And, we can design that the apodization weighted type IDT as an input transducer of the filter and the withdrawal weighted type IDT as an output transducer of the filter from the results of our computer-simulation. Also, we have employed that the number of pairs of the input and output IDT are 2200 pairs and 1000pairs, respectively and used the Kaiser-Bessel window function in order to minimize the effect of ripple. And, while the width and the space of IDT's finger are 6㎛ m and 5.75㎛, respectively and we could obtain the resonable results when the IDT thickness was 6000Å in consideration of the ratio of SAW's wavelength, and IDT aperture is 2mm. Frequency response of the fabricated SAW bandpass filter has the property that the center frequency is about 70MHz, shape factor is less than 1.3, bandwidth at the 1.5dB is probably 1.3MHz, out-band attenuation is almost -45dB, insertion loss is 19dB and ripple in the width of bandpass is 1dB approximately. Therefore, these frequency characteristics of the fabricated SAW bandpass filter are agreed well with the designed values.

• The Korean Journal of Applied Statistics
• /
• v.10 no.1
• /
• pp.69-84
• /
• 1997

We consider a linear calibration problem, $y_i = $$\alpha + \beta (x_i - x_0) + \epsilon_i$, $i=1, 2, {\cdot}{\cdot},n$ $y_f = \alpha + \beta (x_f - x_0) + \epsilon, $ where we observe $(x_i, y_i)$'s for the controlled calibration experiments and later we make inference about $x_f$ from a new observation $y_f$. The objective of the calibration design problem is to find the optimal design $x = (x_i, \cdots, x_n$ that gives the best estimates for $x_f$. We compare Kim(1989)'s Bayesian design which minimizes the expected value of the posterior variance of $x_f$ and some optimal designs from literature. Kim suggested the Bayesian optimal design based on the analysis of the characteristics of the expected loss function and numerical must be equal to the prior mean and that the sum of squares be as large as possible. The designs to be compared are (1) Buonaccorsi(1986)'s AV optimal design that minimizes the average asymptotic variance of the classical estimators, (2) D-optimal and A-optimal design for the linear regression model that optimize some functions of $M(x) = \sum x_i x_i'$, and (3) Hunter & Lamboy (1981)'s reference design from their paper. In order to compare the designs which are optimal in some sense, we consider two criteria. First, we compare them by the expected posterior variance criterion and secondly, we perform the Monte Carlo simulation to obtain the HPD intervals and compare the lengths of them. If the prior mean of $x_f$ is at the center of the finite design interval, then the Bayesian, AV optimal, D-optimal and A-optimal designs are indentical and they are equally weighted end-point design. However if the prior mean is not at the center, then they are not expected to be identical.In this case, we demonstrate that the almost Bayesian-optimal design was slightly better than the approximate AV optimal design. We also investigate the effects of the prior variance of the parameters and solution for the case when the number of experiments is odd.