• 한국전자통신학회논문지
• /
• 제14권2호
• /
• pp.433-438
• /
• 2019

본 논문에서는 기계 서비스 부품 도면에서 숫자를 인식하기 위하여 입력 영상에 대한 전처리와 딥러닝 모델을 제안한다. 서비스 부품 도면의 숫자를 인식하는데 있는 지시선과 도형에 의한 오검출 또는 오인식을 개선하기 위하여 수학적 형태학 필터링 전처리를 한다. 숫자 인식을 위하여 VGG-16 모델을 축소 변형한 7 개의 계층을 가지는 VGG 모델을 적용함으로써 인식 성능을 개선한다. 서비스 부품 도면의 숫자 인식 실험 결과, 제안하는 방법이 인식률 95.57%, 정확도는 92.82%로 종래의 방법에 현저히 개선된 결과를 얻었다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제14권12호
• /
• pp.4664-4681
• /
• 2020

Modulation recognition (MR) plays a key role in cognitive radar, cognitive radio, and some other civilian and military fields. While existing methods can identify the signal modulation type by extracting the signal characteristics, the quality of feature extraction has a serious impact on the recognition results. In this paper, an end-to-end MR method based on long short-term memory (LSTM) and the gated recurrent unit (GRU) is put forward, which can directly predict the modulation type from a sampled signal. Additionally, the sliding window method is applied to fast-changing mixed-modulation signals for which the signal modulation type changes over time. The recognition accuracy on training datasets in different SNR ranges and the proportion of each modulation method in misclassified samples are analyzed, and it is found to be reasonable to select the evenly-distributed and full range of SNR data as the training data. With the improvement of the SNR, the recognition accuracy increases rapidly. When the length of the training dataset increases, the neural network recognition effect is better. The loss function value of the neural network decreases with the increase of the training dataset length, and then tends to be stable. Moreover, when the fast-changing period is less than 20ms, the error rate is as high as 50%. As the fast-changing period is increased to 30ms, the error rates of the GRU and LSTM neural networks are less than 5%.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2007년도 추계학술발표대회
• /
• pp.34-37
• /
• 2007

Face recognition is still a challenging problem in pattern recognition field which is affected by different factors such as facial expression, illumination, pose etc. The facial feature such as eyes, nose, and mouth constitute a complete face. Mouth feature of face is under the undesirable effect of facial expression as many factors contribute the low performance. We proposed a new approach for face recognition under facial expression applying two cascaded classifiers to improve recognition rate. All facial expression images are treated by general purpose classifier at first stage. All rejected images (applying threshold) are used for adaptation using GA for improvement in recognition rate. We apply Gabor Wavelet as a general classifier and Gabor wavelet with Genetic Algorithm for adaptation under expression variance to solve this issue. We have designed, implemented and demonstrated our proposed approach addressing this issue. FERET face image dataset have been chosen for training and testing and we have achieved a very good success.

• 한국정보통신학회:학술대회논문집
• /
• 한국해양정보통신학회 2003년도 춘계종합학술대회
• /
• pp.305-308
• /
• 2003

본 논문에서는 Fuzzy-ART (Fuzzy-Adaptive Resonance Theory) 신경회로망의 패턴인식 성능을 개선하기 위해 Mahalanobis 거리를 이용한 타원형 fuzzy-ART 신경회로망을 제안한다. 제안한 방식은 벡터공간상에서 패턴의 영역을 규정하기 위해 Mahalanobois 거리 개념을 이용한다. 제안한 방식의 유용성을 확인하기 위해 얼굴인식에 적용하였으며, 기존의 방식과 비교 검토한 결과 유용성을 확인하였다.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1995년도 하계학술대회 논문집 B
• /
• pp.931-933
• /
• 1995

In this paper, we develop protype of wierless recognition system using low frequency. Application of this system is very broad. Namely, High-way toll gate, animal management, parking system and industral automation et al. This system is composed of controller, decoder and tag. Controller is personal PC, decoder is signal module and tag is mobile corresponder module. Modulation is ASK, 4,800bps, frequency is 120/60kHz and transmission length is about 80cm. And now we study improvement of stability, low power consumption, compact of tag and transmission length improvement.

• 대한임베디드공학회논문지
• /
• 제3권1호
• /
• pp.1-7
• /
• 2008

In this paper, a MSS-NOVO method that combines the HMM composition method with a noise reduction method is proposed for speech recognition in noisy environments. This combined method starts with noise reduction with modified spectral subtraction (MSS) to enhance the input noisy speech, then the noise and voice composition (NOVO) method is applied for making noise adapted models by using the noise in the non-utterance regions of the enhanced noisy speech. In order to evaluate the effectiveness of our proposed method, we compare MSS-NOVO method with other methods, i.e., SS-NOVO, MWF-NOVO. To set up the noisy speech for test, we add White noise to KLE 452 database with different SNRs range from 0dB to 15dB, at 5dB intervals. From the tests, MSS-NOVO method shows average improvement of 66.5% and 13.6% compared with the existing SS-NOVO method and MWF-NOVO method, respectively. Especially our proposed MSS-NOVO method shows a big improvement at low SNRs.

• International journal of advanced smart convergence
• /
• 제10권2호
• /
• pp.73-78
• /
• 2021

Recently, the number of people who visit the hospital due to diabetes is increasing. According to the Korean Diabetes Association, it is statistically indicated that one in seven adults aged 30 years or older in Korea suffers from diabetes, and it is expected to be more if the pre-diabetes, fasting blood sugar disorders, are combined. In the last study, the validity of Triglyceride and Cholesterol associated with diabetes was confirmed and analyzed using Random Forest. Random Forest has a disadvantage that as the amount of data increases, it uses more memory and slows down the speed. Therefore, in this paper, we compared and analyzed Random Forest and XGBoost, focusing on improvement of learning speed and prevention of memory waste, which are mainly dealt with in machine learning. Using XGBoost, the problem of slowing down and wasting memory was solved, and the accuracy of the diabetes recognition factor was further increased.

• International journal of advanced smart convergence
• /
• 제12권4호
• /
• pp.119-125
• /
• 2023

Smart home in the 4th industrial revolution environment is where all devices in the home are connected to each other to provide the optimal living environment desired by the user. Artificial intelligence speakers are being used as a way to manage and control all devices used in this environment. The function of an artificial intelligence speaker ranges from simple music playback to serving as an interface that controls and manages all devices in a smart home space. In this study, we investigated and analyzed the usability of artificial intelligence speakers based on the current status of domestic and overseas markets and the survey contents of two organizations (Korea Consumer Agency and Korea Information and Communication Policy Institute (KISDI)). In addition, we investigated and analyzed the usability of artificial intelligence speakers. Based on the results of responses from users from two related organizations, major problems were derived, and major improvement measures, such as discovering new functions and improving voice recognition performance, were also described.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2000년도 추계종합학술대회 논문집(4)
• /
• pp.65-68
• /
• 2000

This paper describes an approach for extracting invariant features using a view-based representation and recognizing an object with a high speed search method in FLIR. In this paper, we use a reformulated eigenspace technique based on robust estimation for extracting features which are robust for outlier such as noise and clutter. After extracting feature, we recognize an object using a partial distance search method for calculating Euclidean distance. The experimental results show that the proposed method achieves the improvement of recognition rate compared with standard PCA.

• Communications for Statistical Applications and Methods
• /
• 제11권3호
• /
• pp.503-511
• /
• 2004

In Korea, OCR-based techniques have been developed for digital library construction of historical documents. In this paper, we propose the nonlinear transform of contour directional feature (CDF) vectors using log it and power transforms with skewness criterion to enhance the discriminant power. Experiments were conducted using samples from Seung-jung-won diaries (Diaries of King's Secretaries). Our results show that proposed method outperforms the others like Box-Cox transform in this database.