• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2006.11a
• /
• pp.348-351
• /
• 2006

영상을 통한 감정 인식 기술은 사회의 여러 분야에서 필요성이 대두되고 있음에도 불구하고 인식 과정의 어려움으로 인해 풀리지 않는 문제로 남아 있다. 인간의 움직임을 이용한 감정 인식 기술은 많은 응용이 가능하기 때문에 개발의 필요성이 증대되고 있다. 영상을 통해 감정을 인식하는 시스템은 매우 다양한 기법들이 사용되는 복합적인 시스템이다. 본 논문에서는 이전에 연구된 움직임 추출 방법들을 바탕으로 한 새로운 감정 인식 시스템을 제안한다. 제안된 시스템은 은닉 마르코프 모델을 통해 동정된 분류기를 이용하여 감정을 인식한다. 제안된 시스템의 성능을 평가하기 위해 평가데이터 베이스가 구축되었으며, 이를 통해 제안된 감정 인식 시스템의 성능을 확인하였다.

• International Journal of Computer Science & Network Security
• /
• v.21 no.4
• /
• pp.131-139
• /
• 2021

In current time, anomaly detection is the primary concern of the administrative authorities. Suspicious activity identification is shifting from a human operator to a machine-assisted monitoring in order to assist the human operator and react to an unexpected incident quickly. These automatic surveillance systems face many challenges due to the intrinsic complex characteristics of video sequences and foreground human motion patterns. In this paper, we propose a novel approach to detect anomalous human activity using a hybrid approach of statistical model and Genetic Programming. The feature-set of local motion patterns is generated by a statistical model from the video data in an unsupervised way. This features set is inserted to an enhanced Genetic Programming based classifier to classify normal and abnormal patterns. The experiments are performed using publicly available benchmark datasets under different real-life scenarios. Results show that the proposed methodology is capable to detect and locate the anomalous activity in the real time. The accuracy of the proposed scheme exceeds those of the existing state of the art in term of anomalous activity detection.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.50 no.4
• /
• pp.182-188
• /
• 2013

In this paper, we propose a complex-valued support vector machine (SVM) classifier which process the complex valued signal measured by pulse doppler radar (PDR) to identify moving targets from the background. SVM is widely applied in the field of pattern recognition, but features which used to classify are almost real valued data. Proposed complex-valued SVM can classify the moving target using real valued data, imaginary valued data, and cross-information data. To design complex-valued SVM, we consider slack variables of real and complex axis, and use the KKT (Karush-Kuhn-Tucker) conditions for complex data. Also we apply radial basis function (RBF) as a kernel function which use a distance of complex values. To evaluate the performance of the complex-valued SVM, complex valued data from PDR were classified using real-valued SVM and complex-valued SVM. The proposed complex-valued SVM classification was improved compared to real-valued SVM for dog and human, respectively 8%, 10%, have been improved.

• Journal of KIISE:Software and Applications
• /
• v.31 no.1
• /
• pp.61-70
• /
• 2004

It is difficult to find an appropriate web site because exponentially growing web contains millions of web documents. Personalization of web search can be realized by recommending proper web sites using user profile but more efficient method is needed for estimating preference because user's evaluation on web contents presents many aspects of his characteristics. As user profile has a property of non-linearity, estimation by classifier is needed and combination of classifiers is necessary to anticipate diverse properties. Structure adaptive self-organizing map (SASOM) that is suitable for Pattern classification and visualization is an enhanced model of SOM and might be useful for web mining. Fuzzy integral is a combination method using classifiers' relevance that is defined subjectively. In this paper, estimation of user profile is conducted by using ensemble of SASOM's teamed independently based on fuzzy integral and evaluated by Syskill ＆ Webert UCI benchmark data. Experimental results show that the proposed method performs better than previous naive Bayes classifier as well as voting of SASOM's.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.26 no.1
• /
• pp.87-92
• /
• 2016

In this paper, optimization technique such as particle swarm optimization was used to optimize the parameters of fuzzy Extreme Learning Machine. While the learning speed of conventional neural networks is very slow, that of Extreme Learning Machine is very fast. Fuzzy Extreme Learning Machine is composed of the Extreme Learning Machine with very fast learning speed and fuzzy logic which can represent the linguistic information of the field experts. The general sigmoid function is used for the activation function of Extreme Learning Machine. However, the activation function of Fuzzy Extreme Learning Machine is the membership function which is defined in the procedure of fuzzy C-Means clustering algorithm. We optimize the parameters of the membership functions by using optimization technique such as Particle Swarm Optimization. In order to validate the classification capability of the proposed classifier, we make several experiments with the various machine learning datas.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.10 no.3
• /
• pp.179-187
• /
• 2000

Fractal theory has strengths such as high compression rate and fast decoding time in application to image compression, but it suffers from long comparison time necessary for finding an optimally similar domain block in the encoding stage. This paper proposes a neural network based block classifier which enhances the encoding time significantly by classifying domain blocks into 4 patterns and searching only those blocks having the same pattern with the range block to be encoded. Size of a block is differently determined depending on the image complexity of the block. The proposed algorithm has been tested with three different images having various featrues. The experimental results have shown that the proposed algorithm enhances the compression time by 40% on average compared to the conventional fractal encoding algorithms, while maintaining allowable image qualify of PSNR 30 dB.

• KIPS Transactions on Software and Data Engineering
• /
• v.5 no.2
• /
• pp.89-94
• /
• 2016

This paper proposed a facial expression recognition approach using SIFT feature and SVM classifier. The SIFT was generally employed as feature descriptor at key-points in object recognition fields. However, this paper applied the SIFT descriptor as feature vector for facial expression recognition. In this paper, the facial feature was extracted by applying SIFT descriptor at each sub-block image without key-point detection procedure, and the facial expression recognition was performed using SVM classifier. The performance evaluation was carried out through comparison with binary pattern feature-based approaches such as LBP and LDP, and the CK facial expression database and the JAFFE facial expression database were used in the experiments. From the experimental results, the proposed method using SIFT descriptor showed performance improvements of 6.06% and 3.87% compared to previous approaches for CK database and JAFFE database, respectively.

• Journal of Intelligence and Information Systems
• /
• v.14 no.2
• /
• pp.193-206
• /
• 2008

Medical diagnosis can be considered a classification task which classifies disease types from patient's condition data represented by a set of pre-defined attributes. This study proposes a hybrid genetic algorithm based classification method to develop classifiers for multidimensional pattern classification problems related with medical decision making. The classification problem can be solved by identifying separation boundaries which distinguish the various classes in the data pattern. The proposed method fits a finite number of regional agents to the data pattern by combining genetic algorithms and local adaptive operations. The local adaptive operations of an agent include expansion, avoidance and relocation, one of which is performed according to the agent's fitness value. The classifier system has been tested with well-known medical data sets from the UCI machine learning database, showing superior performance to other methods such as the nearest neighbor, decision tree, and neural networks.

• Journal of KIISE:Software and Applications
• /
• v.37 no.2
• /
• pp.120-128
• /
• 2010

A hypernetwork is a generalized hypo-graph and a probabilistic graphical model based on evolutionary learning. Hypernetwork models have been applied to various domains including pattern recognition and bioinformatics. Nevertheless, conventional hypernetwork models have the limitation that they can manage data with categorical or discrete attibutes only since the learning method of hypernetworks is based on equality comparison of hyperedges with learned data. Therefore, real-valued data need to be discretized by preprocessing before learning with hypernetworks. However, discretization causes inevitable information loss and possible decrease of accuracy in pattern classification. To overcome this weakness, we propose a novel feature-wise L1-distance based method for real-valued attributes in learning hypernetwork models in this study. We show that the proposed model improves the classification accuracy compared with conventional hypernetworks and it shows competitive performance over other machine learning methods.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.47 no.1
• /
• pp.53-61
• /
• 2010

In this paper, we present a color interpolation algorithm that uses novel edge direction estimator and region classifier. The proposed edge direction estimator accurately determines the edge direction based on the correlation between the images obtained by the channel separated and down-sampled Bayer color filter array(CFA) pattern. The correlation is defined based on the similarity between the edge direction in the local region of the image and the shifting direction of the images. Also, the region of an image is defined as the flat, the edge, and the pattern-edge regions, where the edges are appeared repeatedly. When all the pixels in the image are classified into the three different regions, each pixel is interpolated horizontally or vertically according to the estimated direction. Experimental results show that the proposed algorithm outperforms the conventional edge-directed methods on objective and subjective criteria.