• Journal of the Korean Institute of Intelligent Systems
• /
• v.19 no.6
• /
• pp.773-778
• /
• 2009

This paper studied the training methods less affected by the emotional variation for the development of the robust speech recognition system. For this purpose, the effect of emotional variations on the speech signal were studied using speech database containing various emotions. The performance of the speech recognition system trained by using the speech signal containing no emotion is deteriorated if the test speech signal contains the emotions because of the emotional difference between the test and training data. In this study, it is observed that vocal tract length of the speaker is affected by the emotional variation and this effect is one of the reasons that makes the performance of the speech recognition system worse. In this paper, vocal tract normalization method is used to develop the robust speech recognition system for emotional variations. Experimental results from the isolated word recognition using HMM showed that the vocal tract normalization method reduced the error rate of the conventional recognition system by 41.9% when emotional test data was used.

• The Journal of the Acoustical Society of Korea
• /
• v.16 no.2
• /
• pp.71-79
• /
• 1997

In this paper, a codeword-dependent distance normalization(CDDN) and an instar-formed fuzzy smoothing of output distribution are proposed for robust estimation of output probabilities in the FVQ(fuzzy vector quantization)-DHMM(discrete hidden Markov model). The FVQ-DHMM is a variant of DHMM in which the state output probability is estimated by the sum oft he product of the output probability and its weighting factor for each codeword on an input vector. As the performance of the FVQ-DHMM is influenced by weighting factor and output distribution from a state, it is required to get a method to get robust estimation of weighting factors and output distribution for each state. From experimental results, the proposed CDDN method has reduced 24% of error rate over the conventional FVQ-DHMM, and also reduced 79% of error rate when the smoothing of output distribution is also applied to the computation of an output probability. These results indicate that the use of CDDN and the fuzzy smoothing of output distribution to the FVQ-DHMM lead to improved recognition, and therefore it may be used as an alternative to the robust estimation of output probabilities for HMMs.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.7
• /
• pp.2390-2406
• /
• 2022

Scene text recognition has important application value and attracted the interest of plenty of researchers. At present, many methods have achieved good results, but most of the existing approaches attempt to improve the performance of scene text recognition from the image level. They have a good effect on reading regular scene texts. However, there are still many obstacles to recognizing text on low-quality images such as curved, occlusion, and blur. This exacerbates the difficulty of feature extraction because the image quality is uneven. In addition, the results of model testing are highly dependent on training data, so there is still room for improvement in scene text recognition methods. In this work, we present a natural scene text recognizer to improve the recognition performance from the feature level, which contains feature representation and feature enhancement. In terms of feature representation, we propose an efficient feature extractor combined with Representative Batch Normalization and ResNet. It reduces the dependence of the model on training data and improves the feature representation ability of different instances. In terms of feature enhancement, we use a feature enhancement network to expand the receptive field of feature maps, so that feature maps contain rich feature information. Enhanced feature representation capability helps to improve the recognition performance of the model. We conducted experiments on 7 benchmarks, which shows that this method is highly competitive in recognizing both regular and irregular texts. The method achieved top1 recognition accuracy on four benchmarks of IC03, IC13, IC15, and SVTP.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2012.05a
• /
• pp.149-150
• /
• 2012

This paper presents a shot boundary detection method based on the global decision tree that allows for extraction of boundaries of high variations occurring due to camera breaks from frame difference values. For a start, difference values between frames are calculated through local X2-histogram and normalization. Next, the distances between difference values are calculated through normalization.

• Proceedings of the KSRS Conference
• /
• 1999.11a
• /
• pp.271-280
• /
• 1999

The topographical effect resulted from rugged terrains and inhomogeneous spectral characteristics due to the complexly mixed land cover condition of Korea substantially lower the remotely sensed land cover classification accuracy In this study, a topographic correction method using digital elevation model to alleviate the topographic effects. To deal with inhomogeneous spectral characteristic, a hybrid classifier with inclusion of prior probabilities was introduced. This investigation concluded that the topographical normalization and hybrid classification with prior probabilities are effective on rugged landscape. The overall and average classification accuracies were improved by 0.92％ and 1.016％ respectively. The most substantial and noticeable accuracy improvement was observed in forest areas.

• International Journal of Advanced Culture Technology
• /
• v.5 no.2
• /
• pp.54-62
• /
• 2017

Currently, falling to industrial field workers is causing serious injuries. Therefore, many researchers are actively studying the fall by using acceleration sensor, gyro sensor, pressure sensor and image information.Also, as the spread of smartphones becomes common, techniques for determining the fall by using an acceleration sensor built in a smartphone are being studied. The proposed method has complexity due to fusion of various sensor data and it is still insufficient to develop practical application. Therefore, in this paper, we use acceleration sensor module built in smartphone to collect acceleration data, propose a simple falling algorithm based on accelerometer sensor data after normalization and preprocessing, and implement an Android based app.

• Proceedings of the KIPE Conference
• /
• 2002.07a
• /
• pp.300-303
• /
• 2002

With increase of servo motor In industrial and home application, a number of papers related to PMSM control have been researched. Among them, sensorless control schemes are especially concerned in the view point of its cost reduction. In the conventional approach, a rotor position is generally estimated by the integration of estimated rotor speed. In this method, because of their tight relationship between the amplitude of back-emf and rotor position. it is somewhat difficult to find two parameters at the same time. To solve this problem, a novel sensorless control scheme is proposed. It utilizes a back-emf normalization, so it does not requires the variables related with the amplitude of back-emf. The validity of the proposed control scheme was verified through experimental results.

• Journal of Information Science Theory and Practice
• /
• v.6 no.3
• /
• pp.45-60
• /
• 2018

Research performance evaluation in Korean universities follows strict guidelines that specify scoring systems for publication venue categories and formulas for co-authorship credit allocation. To find out how the standards differ across universities and how they differ from bibliometric research evaluation measures, this study analyzed 25 standards from major Korean universities and rankings produced by applying standards and bibliometric measures such as publication and citation counts, normalized impact score, and h-index to the publication data of 195 tenure-track professors of library and information science departments in 35 Korean universities. The study also introduced a novel impact score normalization method to refine the methodology from prior studies. The results showed the university standards to be mostly similar to one another but quite different from citation-driven measures, which suggests the standards are not quite successful in quantifying the quality of research as originally intended.

• The KIPS Transactions:PartB
• /
• v.8B no.4
• /
• pp.389-394
• /
• 2001

본 논문에서는 독립 성분 해석(Independent Component Analysis, ICA) 기법과 인근 평균 및 정규화를 이용한 영상 분류 방법을 제안하였다. ICA에 잡음을 주어 영상을 분류하였을 때, 잡음에 대한 강인성을 증가시키기 위하여, 제안된 인근 평균 및 정규화를 전처리로 적용하였다. 제안된 방법은 전처리 없이 ICA에 주성분 해석(Principal Component Analysis, PCA)을 이용한 것에 비해 잡음에 대한 강인성을 증가시키는 것을 모의 실험을 통하여 확인하였다.

• KSCI Review
• /
• v.15 no.1
• /
• pp.197-201
• /
• 2007

비디오 분할은 샷 경계 검출이라고도 하는데, 비디오를 계층적이고 구조적인 형태로 표현하기 위하여 영상, 문자, 오디오와 같은 매체 속에 포함되어 있는 내용들을 특징별로 분석하여 계층별로 분류하는 작업을 말한다. 본 논문에서는 카메라와 객체의 모션에 보다 강건하고 보다 정확한 결과를 산출하여 충분한 공간 정보를 가지는 지역적 $X^2$-히스토그램 비교 방법을 이용하여 샷 경계를 검출한다. 또한 영상처리에서 영상의 명암 값 향상을 위하여 사용되는 로그함수와 상수를 변형하여 차이 값에 적용하는 정규화 방법을 제시한다. 그리고 샷 경계 검출 알고리즘을 제시하여 일반적인 샷과 갑작스런 샷의 특징을 기반으로 검출한다.