• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2001년도 제14회 신호처리 합동 학술대회 논문집
• /
• pp.503-506
• /
• 2001

본 논문에서는 마이크 입력레벨 조절기의 구현과 이를 이용한 인식률의 향상을 다룬다. 마이크를 통한 음성 입력이 너무 작거나 너무 크면 인식률에 직접 영향을 미치므로 인식에 적합한 입력레벨로 조절할 필요가 있다. 자동 입력레벨 조절기의 구현을 위해 고려할 사항을 연구했으며, 이를 통해 PC환경의 입력레벨 조절기를 구현했다. 수집된 음성 데이터베이스는 켑스트럼 평균차감법(CMS)을 이용하여 채널왜곡을 보상했으며, 구현된 조절기를 이용하여 실험한 결과, 이용하지 않은 경우에 비해 약 50%의 오인식율을 줄일 수 있었다.

• 한국산업정보학회논문지
• /
• 제13권4호
• /
• pp.56-63
• /
• 2008

얼굴 인식률 향상을 위해서는 전처리 단계에서의 영상 보정이 매우 중요하며, 특히 배경 잡음 제거는 얼굴 인식의 정확도에 중대한 영향을 미친다. 본 논문에서는 얼굴 인식률 향상을 위하여 전처리 단계에서 타원 모델을 이용하여 배경 영역을 제거하는 방법을 제안하였다. 사람의 얼굴 윤곽은 타원의 형태를 나타내기 때문에 얼굴 영상에서 타원 모델을 이용할 경우 얼굴 영역을 용이하게 검출할 수 있다. ETRI, ORL, 및 XM2VTS 얼굴 데이터베이스에 대한 실험 분석을 통하여 제안된 방법이 얼굴 인식 성능을 뚜렷하게 개선시켰음을 알 수 있었다.

• 전기전자학회논문지
• /
• 제26권1호
• /
• pp.133-136
• /
• 2022

This paper is about motion gesture recognition system, and proposes the following improvement to the flaws of the current system: a motion gesture recognition system and such algorithm that uses the video image of the entire hand and reading its motion gesture to advance the accuracy of recognition. The motion gesture recognition system includes, an image capturing unit that captures and obtains the images of the area applicable for gesture reading, a motion extraction unit that extracts the motion area of the image, and a hand gesture recognition unit that read the motion gestures of the extracted area. The proposed application of the motion gesture algorithm achieves 20% improvement compared to that of the current system.

• 대한예방한의학회지
• /
• 제7권2호
• /
• pp.45-64
• /
• 2003

This study was conducted to examine the recognition level of the people on oriental medical Services and the need for it's improvement. Data were collected from 1174 residents in Daegu metropolitan city and Gyungbuk province. According to the satisfaction level with each items of oriental medical services, the respondents had positive views on efficacy, kindness, and side-effects. They, however, had negative view on the cost of oriental medical services. In regarding to the priority of improvement of oriental medical system, 'expansion of insurance benefit package' ranked first. Followed by 'safety of herbal medicine(heavy metal/pesticides)', 'improvement of scientific methods and diagnostic technique' etc. For the further development of oriental medical services in the consumer - focused and evidenced-based health care environment, much attention to implement relevant health policy reflecting user's need positively should be made.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2016년도 춘계학술대회
• /
• pp.267-270
• /
• 2016

최근 타이어에는 타이어의 정보를 알 수 있는 정보들이 다양하게 타이어 옆면에 표시되어 있다. 이러한 정보를 이용하는 사람들 및 타이어 관련 회사에서는 타이어의 옆면에 표시된 정보를 가지고 어떤 타이언지 구별을 하게 된다. 타이어의 규격 최대 허용 공기압, 제조일자 등을 일반적으로 사람이 직접 눈으로 봄으로서 이루어지고 있다. 이에 최근 영상 처리 기법을 이용하여 타이어 측면의 돌출 문자 인식을 통한 자동화의 연구가 꾸준히 발표 되고 있지만 문자 인식 및 인식의 개선에 대한 방법이 부족한 실정이다. 또한 기존의 옆면 문자 인식을 영상으로 취득 하는데, 취득시 조명 효과를 적절히 이용하더라도 배경과 문자 부분이 거의 유사간 그레이 레벨 값을 가지게 되어 비교적 분명하지 않은 부분이 많이 산재된다. 본 논문에서는 레이저 센서를 이용한 타이어 옆면 문자를 확인하고 인식, 타이어 옆면의 문자 인식에 관하여 설계를 하였다.

• 한국음향학회:학술대회논문집
• /
• 한국음향학회 1994년도 제11회 음성통신 및 신호처리 워크샵 논문집 (SCAS 11권 1호)
• /
• pp.358-363
• /
• 1994

This paper presents the improvement of the SCHMM using discrete VQ and One-Pass algorithm for keywords recognition in Korean continuous speech. The SCHMM using discrete VQ is a simple model that is composed of a variable mixture gaussian probability density function with dynamic mixture number. One-Pass algorithm is improved such that recognition rates are enhanced by fathoming any undesirable semisyllable with the low likelihood and the high duration penalty, and computation time is reduced by testing only the frame which is dissimilar to the previously testd frame. In recognition experiments for speaker-dependent case, the improved One-Pass algorithm has shown recognition rates as high as 99.7% and has reduced compution time by about 30% compared with the currently abailable one-pass algorithm.

• 말소리와 음성과학
• /
• 제6권2호
• /
• pp.29-34
• /
• 2014

In the Vector Taylor Series (VTS)-based noisy speech recognition methods, Hidden Markov Models (HMM) are usually trained with clean speech. However, better performance is expected by training the HMM with noisy speech. In a previous study, we could find that Minimum Mean Square Error (MMSE) estimation of the training noisy speech in the log-spectrum domain produce improved recognition results, but since the proposed algorithm was done in the log-spectrum domain, it could not be used for the HMM adaptation. In this paper, we modify the previous algorithm to derive a novel mathematical relation between test and training noisy speech in the cepstrum domain and the mean and covariance of the Multi-condition TRaining (MTR) trained noisy speech HMM are adapted. In the noisy speech recognition experiments on the Aurora 2 database, the proposed method produced 10.6% of relative improvement in Word Error Rates (WERs) over the MTR method while the previous MMSE estimation of the training noisy speech produced 4.3% of relative improvement, which shows the superiority of the proposed method.

• 한국컴퓨터정보학회논문지
• /
• 제15권6호
• /
• pp.73-80
• /
• 2010

어휘 인식 시스템은 부정확한 어휘 제공과 유사한 음소 인식으로 인식률이 저하되며 이는 유사한 음소인식 오인식과 효율적 특징 추출 처리를 위한 방법을 필요로 한다. 따라서 본 논문에서는 음소가 갖는 특징을 기반으로 바타챠랴 거리 측정법을 이용한 음소 유사율 오류 보정 개선 시스템을 제안하였다. 음소 유사율은 모노폰으로 훈련시킨 훈련 데이터의 음소에 HMM 특징 추출 방법을 이용하였으며 유사한 음소는 바타챠랴 거리 측정법을 이용하여 정확한 음소로 인식할 수 있도록 유도하여 인식률 향상 효과를 얻을 수 있었다. 이를 유클리디안 거리 측정법과 동적타임 워핑 시스템에 비교한 시스템 성능 평가 결과 1.2%의 향상된 97.91% 인식률을 보였다.

• 한국환경보건학회지
• /
• 제41권6호
• /
• pp.380-388
• /
• 2015

Objectives: Poisonous substances experts engage in policymaking regarding poisonous substances, whereas poisonous substances managers perform legal duties related to poisonous substances management systems at worksites. To understand and improve poisonous substances designation systems, it is necessary to distinguish the roles of these experts and managers. We further aimed to identify problems with poisonous substances appointment systems and provide basic data for their improvement by investigating the knowledge and recognition of the need for system improvement among managers and experts. Methods: To determine the $subjects^{\circ}{\emptyset}$ awareness of poisonous substances designation systems, a selfadministered questionnaire comprised of items on general characteristics and knowledge and recognition of the need for system improvement was issued. Data were analyzed via t-tests and ANOVA using SPSS v 16.0 Results: Regarding system comprehension, managers (n=213; mean 3.28, SD 0.63) significantly differed from experts (n=35; mean 4.08, SD 0.59) (P<0.001). No differences were found in recognition of need for improvement between managers (mean 3.11, SD 0.34) and experts (mean 3.24, SD 0.31). Significant differences in comprehension were found among managers by education level (P<0.05) and company size (P<0.05). Regarding industry type, comprehension was greater in the electrical/electronic and chemical/environmental industries compared to in other industries (P = 0.066). Comprehension was higher among those with five years or more of experience compared to those with less than five years. Regarding recognition of need for improvement, managers showed significant differences by company size (P<0.05). $Experts^{\circ}{\emptyset}$ comprehension showed no differences by general characteristics, while need for improvement differed according to work experience (P<0.001). Conclusion: The study found that the bigger the company size, the higher the level of comprehension by managers and the experts. In the case of both groups, where a higher comprehension of the poisonous substances designation system was evident, they recognized the necessity of its improvement.

• 대한음성학회지:말소리
• /
• 제44호
• /
• pp.105-113
• /
• 2002

In Korean, each digit is monosyllable and some pairs are known to have high confusability, causing performance degradation of connected digit recognition systems. To improve the performance, in this paper, we employ various discriminant analyses (DA) including Linear DA (LDA), Weighted Pairwise Scatter LDA WPS-LDA), Heteroscedastic Discriminant Analysis (HDA), and Maximum Likelihood Linear Transformation (MLLT). We also examine several combinations of various DA for additional performance improvement. Experimental results show that applying any DA mentioned above improves the string accuracy, but the amount of improvement of each DA method varies according to the model complexity or number of mixtures per state. Especially, more than 20% of string error reduction is achieved by applying MLLT after WPS-LDA, compared with the baseline system, when class level of DA is defined as a tied state and 1 mixture per state is used.