• 전자공학회논문지B
• /
• 제33B권8호
• /
• pp.105-113
• /
• 1996

In this paper, a new korean digita speech recognition technique was proposed using muktolayer perceptron (MLP). In spite of its weakness in dynamic signal recognition, MLP was adapted for this model, cecause korean syllable could give static features. It is so simle in its structure and fast in its computing that MLP was used to the suggested system. MLP's input vectors was transformed using karhunen-loeve transformation (KLT), which compress signal successfully without losin gits separateness, but its physical properties is changed. Because the suggested technique could extract static features while it is not affected from the changes of syllable lengths, it is effectively useful for korean numeric recognition system. Without decreasing classification rates, we can save the time and memory size for computation using KLT. The proposed feature extraction technique extracts same size of features form the tow same parts, front and end of a syllable. This technique makes frames, where features are extracted, using unique size of windows. It could be applied for continuous speech recognition that was not easy for the normal neural network recognition system.

• 제어로봇시스템학회논문지
• /
• 제5권7호
• /
• pp.821-826
• /
• 1999

This paper proposes a face recognition method using feature spaces and image vectors in the image plane. We obtain the 2-D feature space using the self-organizing map which has two inputs from the axis of the given image. The image vector consists of its weights and the average gray levels in the feature space. Also, we can reconstruct an normalized face by using the image vector having no connection with the size of the given face image. In the proposed method, each face is recognized with the best match of the feature spaces and the maximum match of the normally retrieval face images, respectively. For enhancing recognition rates, our method combines the two recognition methods by the feature spaces and the retrieval images. Simulations are conducted on the ORL(Olivetti Research laboratory) images of 40 persons, in which each person has 10 facial images, and the result shows 100% recognition and 14.5% rejection rates for the 20$\times$20 feature sizes and the 24$\times$28 retrieval image size.

• 한국전자통신학회논문지
• /
• 제14권6호
• /
• pp.1207-1214
• /
• 2019

4차 산업혁명의 기술이 우리도 모르는 사이 우리의 삶 속으로 스며들고 있다. CNN이 이미지 인식 분야에서 탁월한 능력을 보여준 이후 많은 IoT 기반 홈보안 시스템은 침입자로부터 가족과 가정을 보호하며 얼굴을 인식하기 위한 좋은 생체인식 방법으로 CNN을 사용하고 있다. 본 논문에서는 2D와 2.5D 이미지에 대하여 여러 종류의 입력 이미지 크기와 필터를 가지고 있는 CNN의 구조를 연구한다. 실험 결과는 50*50 크기를 가진 2.5D 입력 이미지, 2 컨벌류션과 맥스풀링 레이어, 3*3 필터를 가진 CNN 구조가 0.966의 인식률을 보여 주었고, 1개의 입력 이미지에 대하여 가장 긴 CPU 소비시간은 0.057S로 나타났다. 홈보안 시스템은 좋은 얼굴 인식률과 짧은 연산 시간을 요구하므로 본 논문에서 제안한 구조의 CNN은 홈보안 시스템에서 얼굴인식을 기반으로 하는 액추에이터 제어 등에 적합한 방법이 될 것이다.

• 한국의상디자인학회지
• /
• 제9권3호
• /
• pp.87-97
• /
• 2007

This study is to investigate the KS size recognition and mother's preference of female children's apparel. The practical research is performed for 150 mothers lived in Seoul and are randomly selected to their age, female children's number, education and income level. For statistical analysis and evaluation of survey data, frequency and percentage use contingency table. Findings in this study as follow: 1. Mother's preference for purchasing the girl's garments shows the significant differences of their subject characteristics such as age, girl's number, education and income level. 2. Mother's recognition about KSK 9403: 2004 sizing system for girl's garments does not show the significant differences of their subject properties. Most mothers only know the part of the KS size specifications because KS sizing systems are complex. So KS sizing systems must be simplified and respecified to understand the KS for mothers easily when purchasing their girl's garments. In summary this paper investigates mother's preference and recognition about KS sizing system for the girl's garments.

• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제3권3호
• /
• pp.135-138
• /
• 2014

본 논문에서는 2D-DCT와 EHMM 알고리즘을 이용하여 과적합에 강인한 얼굴 표정인식 방법을 고안하였다. 특히, 본 논문에서는 2D-DCT 특징추출을 위한 윈도우 크기를 크게 설정하여 EHMM의 관측벡터를 추출함으로써, 표정인식 성능 향상을 도모하였다. 제안 방법의 성능평가는 공인 CK 데이터베이스와 JAFFE 데이터베이스를 이용하여 수행되었고, 실험 결과로부터 특징추출 윈도우의 크기가 커질수록 표정 인식률이 향상됨을 확인하였다. 또한, CK 데이터베이스를 이용하여 표정 모델을 생성하고 JAFFE 데이터베이스 전체 샘플을 테스트한 결과, 제안 방법은 87.79%의 높은 인식률을 보였으며, 기존의 히스토그램 특징 기반의 표정인식 접근법보다 46.01~50.05%의 향상된 인식률을 보였다.

• 전자공학회논문지CI
• /
• 제41권5호
• /
• pp.33-40
• /
• 2004

본 논문에서는 웨이블릿 압축을 이용하여 얼굴 데이터베이스를 구축하고, 주성분 분석(Principal Component Analysis : PCA) 알고리듬을 이용하여 얼굴 인식률을 비교한다. 일반적인 얼굴인식 방법은 정규화된 크기를 이용하여 데이터베이스를 구축하고, 얼굴 인식을 한다. 제안된 방법은 정규화된 크기(92×112)의 영상을 웨이블릿 압축으로 1단계, 2단계, 3단계로 변환하고 데이터베이스를 구축한다. 입력 영상도 웨이블릿으로 압축하고 PCA 알고리듬으로 얼굴인식 실험을 하였다 실험을 통하여 제안된 방법은 기존 얼굴영상의 정보를 축소할 뿐만 아니라 처리속도도 향상되었다. 또한 제안된 방법은 원본 영상이 99.05%, 1단계 99.05%, 2단계 98.93%, 3단계 98.54% 정도의 인식률을 보였으며, 대량의 얼굴 데이터베이스를 구축하여 얼굴인식을 하는데 가능함을 보였다.

• 말소리와 음성과학
• /
• 제1권2호
• /
• pp.3-8
• /
• 2009

In an effort to improve speech recognition, we investigated the effect of the number of clusters. In usual LBG clustering, the number of codebook clusters is doubled on each bifurcation and hence cannot be chosen arbitrarily in a natural way. To have the number of clusters at our control, we combined adaptive resonance theory (ART2) with LBG and perform the clustering in two stages. The codebook thus formed was used in subsequent processing of fuzzy vector quantization (FVQ) and HMM for speech recognition tests. Compared to conventional LBG, our method was shown to reduce the best recognition error rate by 0${\sim$}0.9% depending on the vocabulary size. The result also showed that between 400 and 800 would be the optimal number of clusters in the limit of small and large vocabulary speech recognitions of isolated words, respectively.

• 전자공학회논문지B
• /
• 제30B권2호
• /
• pp.8-18
• /
• 1993

A multi-font, multi-size and high speed character recognition system is designed. The design principles are simpilcity of algorithm, adaptibility, learnability, hierachical data processing and attention by feed back. For the multi-size character recognition, the extracted character images are normalized. A hierachical classifier classifies the feature vectors. Feature is extracted by applying the directional receptive field after the directional dege filter processing. The hierachical classifier is consist of two pre-classifiers and one decision making classifier. The effect of two pre-classifiers is prediction to the final decision making classifier. With the pre-classifiers, the time to compute the distance of the final classifier is reduced. Recognition rate is 95% for the three documents printed in three kinds of fonts, total 1,700 characters. For high speed implemention, a multiprocessor system with the ring structure of four transputers is implemented, and the recognition speed of 30 characters per second is aquired.

• 전자공학회논문지B
• /
• 제33B권8호
• /
• pp.30-41
• /
• 1996

This paper presents a comparative study of four pattern recognition algorithms which are invariant to translations, rotations, and scale changes of the input object; namely, object shape features (OSF), geometrica fourier mellin transform (GFMT), moment invariants (MI), and centered polar exponential transform (CPET). Pattern description is obviously one of the most important aspects of pattern recognition, which is useful to describe the object shape independently of translation, rotation, or size. We first discuss problems that arise in the conventional invariant pattern recognition algorithms, or size. We first discuss problems that arise in the coventional invariant pattern recognition algorithms, then we analyze their performance using the same criterion. Computer simulations with several distorted images show that the CPET algorithm yields better performance than the other ones.

• International journal of advanced smart convergence
• /
• 제8권3호
• /
• pp.95-101
• /
• 2019

Recently, there have been many issues related to text recognition using Tesseract. One of these issues is that the text recognition accuracy is significantly lower for smaller fonts. Tesseract extracts text by creating an outline with direction in the image. By searching the Tesseract database, template matching with characters with similar feature points is used to select the character with the lowest error. Because of the poor text extraction, the recognition accuracy is lowerd. In this paper, we compared text recognition accuracy after applying various super-resolution methods to smaller text images and experimented with how the recognition accuracy varies for various image size. In order to recognize small Korean text images, we have used super-resolution algorithms based on deep learning models such as SRCNN, ESRCNN, DSRCNN, and DCSCN. The dataset for training and testing consisted of Korean-based scanned images. The images was resized from 0.5 times to 0.8 times with 12pt font size. The experiment was performed on x0.5 resized images, and the experimental result showed that DCSCN super-resolution is the most efficient method to reduce precision error rate by 7.8%, and reduce the recall error rate by 8.4%. The experimental results have demonstrated that the accuracy of text recognition for smaller Korean fonts can be improved by adding super-resolution methods to the OCR preprocessing module.