• Journal of information and communication convergence engineering
• /
• v.5 no.3
• /
• pp.245-248
• /
• 2007

In this paper, we research the content-based analysis and classification according to the composition of the feature parameter data base for the audio data to implement the audio data index and searching system. Audio data is classified to the primitive various auditory types. We described the analysis and feature extraction method for the feature parameters available to the audio data classification. And we compose the feature parameters data base in the index group unit, then compare and analyze the audio data centering the including level around and index criterion into the audio categories. Based on this result, we compose feature vectors of audio data according to the classification categories, and simulate to classify using discrimination function.

• Transactions of the Korean Society of Machine Tool Engineers
• /
• v.16 no.6
• /
• pp.62-70
• /
• 2007

In this study, the natural flaws in welding parts are classified using the signal pattern classification method. The storage digital oscilloscope including FFT function and enveloped waveform generator is used and the signal pattern recognition procedure is made up the digital signal processing, feature extraction, feature selection and classifier design. It is composed with and discussed using the distance classifier that is based on euclidean distance the empirical Bayesian classifier. Feature extraction is performed using the class-mean scatter criteria. The signal pattern classification method is applied to the signal pattern recognition of natural flaws.

• Journal of Information Processing Systems
• /
• v.18 no.6
• /
• pp.794-802
• /
• 2022

The recovery of reasonable depth information from different scenes is a popular topic in the field of computer vision. For generating depth maps with better details, we present an efficacious monocular depth prediction framework with coordinate attention and feature fusion. Specifically, the proposed framework contains attention, multi-scale and feature fusion modules. The attention module improves features based on coordinate attention to enhance the predicted effect, whereas the multi-scale module integrates useful low- and high-level contextual features with higher resolution. Moreover, we developed a feature fusion module to combine the heterogeneous features to generate high-quality depth outputs. We also designed a hybrid loss function that measures prediction errors from the perspective of depth and scale-invariant gradients, which contribute to preserving rich details. We conducted the experiments on public RGBD datasets, and the evaluation results show that the proposed scheme can considerably enhance the accuracy of depth prediction, achieving 0.051 for log10 and 0.992 for δ<1.25³ on the NYUv2 dataset.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.40 no.4
• /
• pp.266-274
• /
• 2003

Distance Fourier descriptors of local target boundary and feature information fusion using multiple MLPs (Multilayer perceptrons) are proposed. They are used to identify nonoccluded and partially occluded targets in natural FLIR (forward-looking infrared) images. After segmenting a target, radial Fourier descriptors as global shape features are defined from the target boundary. A target boundary is partitioned into four local boundaries to extract local shape features. In a local boundary, a distance function is defined from boundary points and a line between two extreme points. Distance Fourier descriptors as local shape features are defined by using distance function. One global feature vector and four local feature vectors are used as input data for multiple MLPs to determine final identification result of the target. In the experiments, we show that the proposed method is superior to the traditional feature sets with respect to the identification performance.

• Journal of KIISE:Software and Applications
• /
• v.33 no.8
• /
• pp.692-702
• /
• 2006

In this study, a weighted fuzzy min-max (WFMM) neural network model for pattern classification is proposed. The model has a modified structure of FMM neural network in which the weight concept is added to represent the frequency factor of feature values in a learning data set. First we present in this paper a new activation function of the network which is defined as a hyperbox membership function. Then we introduce a new learning algorithm for the model that consists of three kinds of processes: hyperbox creation/expansion, hyperbox overlap test, and hyperbox contraction. A weight adaptation rule considering the frequency factors is defined for the learning process. Finally we describe a feature analysis technique using the proposed model. Four kinds of relevance factors among feature values, feature types, hyperboxes and patterns classes are proposed to analyze relative importance of each feature in a given problem. Two types of practical applications, Fisher's Iris data and Cleveland medical data, have been used for the experiments. Through the experimental results, the effectiveness of the proposed method is discussed.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.39 no.4
• /
• pp.370-380
• /
• 2002

This paper proposes an ATR (Automatic Target Recognition) algorithm for identifying non-occluded and occluded military vehicles in natural FLIR (Forward Looking InfraRed) images. After segmenting a target, a radial function is defined from the target boundary to extract global shape features. Also, to extract local shape features of upper region of a target, a distance function is defined from boundary points and a line between two extreme points. From two functions and target contour, four global and four local shape features are proposed. They are much more invariant to translation, rotation and scale transform than traditional feature sets. In the experiments, we show that the proposed feature set is superior to the traditional feature sets with respect to the similarity-transform invariance and recognition performance.

• Journal of Broadcast Engineering
• /
• v.24 no.2
• /
• pp.234-242
• /
• 2019

In this paper, we propose a deep neural network which extracts SIFT feature points by determining whether the center pixel of a cropped image is a SIFT feature point. The data set of this network consists of a DIV2K dataset cut into $33{\times}33$ size and uses RGB image unlike SIFT which uses black and white image. The ground truth consists of the RobHess SIFT features extracted by setting the octave (scale) to 0, the sigma to 1.6, and the intervals to 3. Based on the VGG-16, we construct an increasingly deep network of 13 to 23 and 33 convolution layers, and experiment with changing the method of increasing the image scale. The result of using the sigmoid function as the activation function of the output layer is compared with the result using the softmax function. Experimental results show that the proposed network not only has more than 99% extraction accuracy but also has high extraction repeatability for distorted images.

• Journal of the Optical Society of Korea
• /
• v.18 no.4
• /
• pp.317-329
• /
• 2014

In vision measurement systems based on structured light, the key point of detection precision is to determine accurately the central position of the projected laser line in the image. The purpose of this research is to extract laser line centers based on a decision function generated to distinguish the real centers from candidate points with a high recognition rate. First, preprocessing of an image adopting a difference image method is conducted to realize image segmentation of the laser line. Second, the feature points in an integral pixel level are selected as the initiating light line centers by the eigenvalues of the Hessian matrix. Third, according to the light intensity distribution of a laser line obeying a Gaussian distribution in transverse section and a constant distribution in longitudinal section, a normalized model of Hessian matrix eigenvalues for the candidate centers of the laser line is presented to balance reasonably the two eigenvalues that indicate the variation tendencies of the second-order partial derivatives of the Gaussian function and constant function, respectively. The proposed model integrates a Gaussian recognition function and a sinusoidal recognition function. The Gaussian recognition function estimates the characteristic that one eigenvalue approaches zero, and enhances the sensitivity of the decision function to that characteristic, which corresponds to the longitudinal direction of the laser line. The sinusoidal recognition function evaluates the feature that the other eigenvalue is negative with a large absolute value, making the decision function more sensitive to that feature, which is related to the transverse direction of the laser line. In the proposed model the decision function is weighted for higher values to the real centers synthetically, considering the properties in the longitudinal and transverse directions of the laser line. Moreover, this method provides a decision value from 0 to 1 for arbitrary candidate centers, which yields a normalized measure for different laser lines in different images. The normalized results of pixels close to 1 are determined to be the real centers by progressive scanning of the image columns. Finally, the zero point of a second-order Taylor expansion in the eigenvector's direction is employed to refine further the extraction results of the central points at the subpixel level. The experimental results show that the method based on this normalization model accurately extracts the coordinates of laser line centers and obtains a higher recognition rate in two group experiments.

• The Journal of the Acoustical Society of Korea
• /
• v.24 no.7
• /
• pp.410-415
• /
• 2005

Recently, most of the speaker verification systems are based on the pattern recognition approach method. And performance of the pattern-classifier depends on how to classify a variety of speakers' feature parameters. In order to classify feature parameters efficiently and effectively, it is of great importance to enlarge variations between speakers and effectively measure distances between feature parameters. Therefore, this paper would suggest the positively mixed model scheme that can enlarge inter-speaker variation by searching the individual model with world model at the same time. During decision procedure, we can maximize inter-speaker variation by using the proposed mixed model scheme. We also make use of a symbol probability weighting function in this system so as to reduce vector quantization errors by measuring symbol probability derived from the distance rate of between the world codebook and individual codebook. As the result of our experiment using this method, we could halve the Detection Cost Function (DCF) of the system from $2.37\%\;to\;1.16\%$.

• Proceedings of the Korean Society for Agricultural Machinery Conference
• /
• 2017.04a
• /
• pp.166-166
• /
• 2017

모돈의 건강 상태를 정량 지수화 하기 위한 연구를 수행 중이다. 지제이상, 섭식 불량, 수면 패턴 등의 운동 특성 분석을 위하여 복수의 초음파 센서를 이용하였다. 시계열 계측 신호를 분석하여 정량 지수화를 수행하는 과정에서 주파수 도메인 분석을 시도하였다. 이 과정에서 주파수 도메인의 분해능에 따른 편차 극복을 위한 비선형 모델링을 수행하였다. 또한 인접한 시계열 데이터 구간 간의 상관성 분석이 가능하면 대용량 데이터의 실시간 처리로 인한 지연 시간 극복 및 기대되는 예후에 대한 조기 진단이 가능할 것이다. 본 연구에서는 구글에서 제공하는 Tensorflow와 NVIDIA에서 제공하는 CUDA 엔진을 동시 적용한 심층 학습 시스템을 이용하였다. 전 처리를 위하여 주파수 분해능 (2분, 3분, 5분, 7분, 11분, 13분, 17분, 19분)에 따른 데이터 집합을 1단계로 두고, 상위 10 순위 안에 드는 파워 스펙트럼 밀도의 크기를 2단계로 하여, 총 2~10개의 입력 노드를 순차적으로 선정하였고, 동일한 방식으로 인접한 시계열의 파워 스펙터럼 밀도를 순위를 변화시켜 지정하였다. 대표적인 심층학습 모델인 Softmax regression with a multilayer convolutional network를 이용하여 Recursive feature selection 경우의 수를 $8{\times}9{\times}9$로 총 648 가지 선정하고, Epoch는 10,000회로 지정하였다. Calibration 모델링의 경우 Cost function이 10% 이하인 경우 해당 경우의 학습을 중단하였으며, 모델 간 상호 교차 검증을 수행하기 위하여 $_8C_2{\times}_8C_2{\times}_8C_2$ 경우의 수에 대한 Verification test를 수행하였다. Calibration 과정 상 모든 경우에 대하여 10% 이하의 Cost function 값을 보였으나, 검증 테스트 과정에서 모든 경우에 대하여 $r^2$ < 0.5 인 결정 계수 값이 나타났다. 단적으로 심층학습 모델의 과도한 적합(Over fitting) 방식의 한계를 보인 것이라 판단할 수 있다. 적합한 Feature selection 및 심층 학습 모델에 대한 지속적이고 추가적인 고려를 통해 과도적합을 해소함과 동시에 실효적이고 활용 가능한 Classification을 위한 입, 출력 노드 단의 전후 Indexing, Quantization에 대한 고려가 필요할 것이다. 이를 통해 모돈 생체 정보 정량화를 위한 지능형 현장 진단 기술 연구를 지속할 것이다.