• 한국융합학회논문지
• /
• 제7권4호
• /
• pp.1-7
• /
• 2016

구조적 MRI 영상은 여러 단 변량과 다변량 방법을 위해 그레이 메터 (GM), 화이트 메터 (WM), 뇌척수액 (CSF) 세션화 과정을 하고 난후 형태계측학적 특징을 추출하기 위해 사용한다. 새로운 접근 방법은 매우 가벼운 알츠하이머 병에서 가벼운 알츠하이머병의 진단을 위해 적용된다. 간이정신상태검사에 따른 형태계측학적 특징과 가우시안 복합 모델 파라미터를 결합하여 정상인으로부터 알츠하이머 병 환자로 분류하는 방법을 제안한다. 결합한 특징은 주성분 분석 기법을 이용한 고차원의 저주를 제거한 후 다중 커널 SVM 분류기에 공급한다. 제안한 진단 방법의 실험적 결과는 90%이상의 특성도와 고민감도에 따라 다중 커널 SVM을 가진 층화 정확도가 96%까지 최대 산출한다.

• 한국음향학회지
• /
• 제39권2호
• /
• pp.131-136
• /
• 2020

음성 인식기를 대기모드에서 동작 모드로 전환하기 위해 발화하는 짧은 단어를 기동어(Wake Up Word, WUW)라고 하며, 음성 인식기를 실제로 사용하는 사용자가 지정한 기동어를 사용자 정의 기동어라고 한다. 본 논문에서는 이러한 사용자 정의 기동어를 인식하기 위해 기존의 Gaussian Mixture Model-Hidden Markov Model(GMM-HMM) 기반의 시스템, Linear Discriminant Analysis(LDA)를 적용한 LDA-GMM-HMM 기반의 시스템과, LDA-GMM-HMM 모델에서 GMM을 Deep Neural Network(DNN)로 대체한 LDA-DNN-HMM 기반의 시스템을 제작하고 각 시스템의 사용자 정의 기동어 인식 성능 및 비기동어 거절 성능을 비교한다. 또한 기동어 인식기의 체감 성능을 향상시키고자 각 모델에 threshold를 적용하여 기동어 인식 실패율을 약 10 % 수준으로 감소 시킨 후에 비기동어(non-WUW)의 거절 실패율을 비교 평가한다. Threshold 적용시에 LDA-DNN-HMM 기반의 시스템의 경우 기동어 인식 실패율 9.84 % 수준에서 비기동어 거절 실패율이 0.0058 %의 인식 성능을 나타내어 LDA-GMM-HMM 시스템 보다 약 4.82배 향상된 비기동어 거절 성능을 나타낸다. 이러한 결과는 본 논문에서 제작한 LDA-DNN-HMM 모델이 사용자 정의 기동어 인식 시스템을 구축하는데 효과적임을 입증한다.

• 음성과학
• /
• 제10권2호
• /
• pp.35-43
• /
• 2003

General speaker verification systems improve their recognition performances by normalizing log likelihood ratio, using a speaker model and its background speaker model that are required to be verified. So these systems rely heavily on the availability of much speaker independent databases for background speaker model design. This constraint, however, may be a burden in practical and portable devices such as palm-top computers or wireless handsets which place a premium on computations and memory. In this paper, new approach for the GMM-based background model design used in portable speaker verification system is presented when the enrollment data is available. This approach is to modify three parameters of GMM speaker model such as mixture weights, means and covariances along with reduced mixture order. According to the experiment on a 20 speaker population from YOHO database, we found that this method had a promise of effective use in a portable speaker verification system.

• 한국콘텐츠학회:학술대회논문집
• /
• 한국콘텐츠학회 2016년도 춘계 종합학술대회 논문집
• /
• pp.11-12
• /
• 2016

SVM (Support Vector machine) is powerful machine-learning method, and obtains better performance than traditional methods in the applications of muti-dimension nonlinear pattern classification. For the case of SVM model training and low efficiency in large samples, this paper proposes a combination of statistical parameters of the GMM-UBM (Universal Background Model) model. It is very effective to solve the problem of the large sample for the SVM training. The experiment is carried on four special dynamic hand gestures using the EPIC sensors. And the results show that the improved dynamic hand gesture recognition system has a high recognition rate up to 96.75%.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2006년도 하계종합학술대회
• /
• pp.245-246
• /
• 2006

In this paper, we proposed using MFCC coefficients(Mel-Scaled Cepstral Coefficients) and a simple but efficient classifying method. Many other features: IAV, zero crossing, LPCC, $\ldot$ and their derivatives are also tested and compared with MFCC coefficients in order to find the best combination. GMM and HMM (Discrete and Continuous Hidden Markov Model), are studied as well in the hope that the use of continuous distribution and the temporal evolution of this set of features will improve the quality of emotion recognition.

• ETRI Journal
• /
• 제33권6호
• /
• pp.914-923
• /
• 2011

This paper concerns robust and reliable speaker model training for text-independent speaker verification. The baseline speaker modeling approach is the Gaussian mixture model (GMM). In text-independent speaker verification, the amount of speech data may be different for speakers. However, we still wish the modeling approach to perform equally well for all speakers. Besides, the modeling technique must be least vulnerable against unseen data. A traditional approach for GMM training is expectation maximization (EM) method, which is known for its overfitting problem and its weakness in handling insufficient training data. To tackle these problems, variational approximation is proposed. Variational approaches are known to be robust against overtraining and data insufficiency. We evaluated the proposed approach on two different databases, namely KING and TFarsdat. The experiments show that the proposed approach improves the performance on TFarsdat and KING databases by 0.56% and 4.81%, respectively. Also, the experiments show that the variationally optimized GMM is more robust against noise and the verification error rate in noisy environments for TFarsdat dataset decreases by 1.52%.

• 융합신호처리학회 학술대회논문집
• /
• 한국신호처리시스템학회 2003년도 하계학술대회 논문집
• /
• pp.18-21
• /
• 2003

In general it has been considered to be the difficult problem that we divide continuous speech into short interval with having identical phoneme quality. In this paper we used Gaussian Mixture Model (GMM) related to probability density to divide speech into phonemes, an initial, medial, and final sound. From them we peformed continuous speech recognition. Decision boundary of phonemes is determined by algorithm with maximum frequency in a short interval. Recognition process is performed by Continuous Hidden Markov Model(CHMM), and we compared it with another phoneme divided by eye-measurement. For the experiments result we confirmed that the method we presented is relatively superior in auto-segmentation in korean speech.

• Journal of Information Processing Systems
• /
• 제14권2호
• /
• pp.539-551
• /
• 2018

Images are unavoidably contaminated with different types of noise during the processes of image acquisition and transmission. The main forms of noise are impulse noise (is also called salt and pepper noise) and Gaussian noise. In this paper, an effective method of removing mixed noise from images is proposed. In general, different types of denoising methods are designed for different types of noise; for example, the median filter displays good performance in removing impulse noise, and the wavelet denoising algorithm displays good performance in removing Gaussian noise. However, images are affected by more than one type of noise in many cases. To reduce both impulse noise and Gaussian noise, this paper proposes a denoising method that combines adaptive median filtering (AMF) based on impulse noise detection with the wavelet threshold denoising method based on a Gaussian mixture model (GMM). The simulation results show that the proposed method achieves much better denoising performance than the median filter or the wavelet denoising method for images contaminated with mixed noise.

• International journal of advanced smart convergence
• /
• 제1권2호
• /
• pp.39-42
• /
• 2012

This paper proposed a method that detects emergency situations in a video stream using MHI (Motion History Image) and template matching for a video-based intelligent security system. The proposed method creates a MHI of each human object through image processing technique such as background removing based on GMM (Gaussian Mixture Model), labeling and accumulating the foreground images, then the obtained MHI is compared with the existing MHI templates for detecting an emergency situation. To evaluate the proposed emergency detection method, a set of experiments on the dataset of video clips captured from a security camera has been conducted. And we successfully detected emergency situations using the proposed method. In addition, the implemented system also provides MMS (Multimedia Message Service) so that a security manager can deal with the emergency situation appropriately.

• 제어로봇시스템학회논문지
• /
• 제18권3호
• /
• pp.285-294
• /
• 2012

Using bearing measurement only, target motion state is not directly obtained so that TMA (Target Motion Analysis) is needed for this situation. TMA is a nonlinear estimation technique used in passive SONAR systems. Also it is the one of important techniques for underwater combat management systems. TMA can be divided to two parts: batch estimation and sequential estimation. It is preferable to use sequential estimation for reducing computational load as well as adaptively to target maneuvers, batch estimation is still required to attain target initial state vector for convergence of sequential estimation. Selection of batch time interval which depends on observability is critical in TMA performance. Batch estimation in general utilizes predetermined batch time interval. In this paper, we propose a new method called the BTIS (Batch Time Interval and Initial State Estimation). The proposed BTIS estimates target initial status and determines the batch time interval sequentially by using a bank of GMM-TS (Gaussian Mixture Measurement-Track Splitting) filters. The performance of the proposal method is verified by a Monte Carlo simulation study.