• Journal of information and communication convergence engineering
• /
• v.8 no.2
• /
• pp.185-190
• /
• 2010

In this paper we propose the CBIR algorithm which is based on a novel image block method that combined both color and edge feature. The main drawback of global histogram representation is dependent of the color without spatial or shape information, a new image block method that divided the image to 8 related blocks which contained more information of the image is utilized to extract image feature. Based on these 8 blocks, histogram equalization and edge detection techniques are also used for image retrieval. The experimental results show that the proposed image block method has better ability of characterizing the image contents than traditional block method and can perform the retrieval system efficiently.

• ETRI Journal
• /
• v.35 no.1
• /
• pp.100-108
• /
• 2013

In this paper, a feature extraction (FE) method is proposed that is comparable to the traditional FE methods used in automatic speech recognition systems. Unlike the conventional spectral-based FE methods, the proposed method evaluates the similarities between an embedded speech signal and a set of predefined speech attractor models in the reconstructed phase space (RPS) domain. In the first step, a set of Gaussian mixture models is trained to represent the speech attractors in the RPS. Next, for a new input speech frame, a posterior-probability-based feature vector is evaluated, which represents the similarity between the embedded frame and the learned speech attractors. We conduct experiments for a speech recognition task utilizing a toolkit based on hidden Markov models, over FARSDAT, a well-known Persian speech corpus. Through the proposed FE method, we gain 3.11% absolute phoneme error rate improvement in comparison to the baseline system, which exploits the mel-frequency cepstral coefficient FE method.

• ETRI Journal
• /
• v.34 no.4
• /
• pp.572-582
• /
• 2012

The discrete Gaussian-Hermite moment (DGHM) is a global feature representation method that can be applied to square images. We propose a modified DGHM (MDGHM) method and an MDGHM-based scale-invariant feature transform (MDGHM-SIFT) descriptor. In the MDGHM, we devise a movable mask to represent the local features of a non-square image. The complete set of non-square image features are then represented by the summation of all MDGHMs. We also propose to apply an accumulated MDGHM using multi-order derivatives to obtain distinguishable feature information in the third stage of the SIFT. Finally, we calculate an MDGHM-based magnitude and an MDGHM-based orientation using the accumulated MDGHM. We carry out experiments using the proposed method with six kinds of deformations. The results show that the proposed method can be applied to non-square images without any image truncation and that it significantly outperforms the matching accuracy of other SIFT algorithms.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.7 no.11
• /
• pp.2824-2838
• /
• 2013

There has been a lot of attention paid recently to analyze dynamic human gestures that vary over time. Most attention to dynamic gestures concerns with spatio-temporal features, as compared to analyzing each frame of gestures separately. For accurate dynamic gesture recognition, motion feature extraction algorithms need to find representative features that uniquely identify time-varying gestures. This paper proposes a new feature-extraction algorithm using temporal self-similarity based on a hierarchical human model. Because a conventional temporal self-similarity method computes a whole movement among the continuous frames, the conventional temporal self-similarity method cannot recognize different gestures with the same amount of movement. The proposed model-based temporal self-similarity method groups body parts of a hierarchical model into several sets and calculates movements for each set. While recognition results can depend on how the sets are made, the best way to find optimal sets is to separate frequently used body parts from less-used body parts. Then, we apply a multiclass support vector machine whose optimization algorithm is based on structural support vector machines. In this paper, the effectiveness of the proposed feature extraction algorithm is demonstrated in an application for taebo gesture recognition. We show that the model-based temporal self-similarity method can overcome the shortcomings of the conventional temporal self-similarity method and the recognition results of the model-based method are superior to that of the conventional method.

• Journal of Information Processing Systems
• /
• v.16 no.6
• /
• pp.1293-1308
• /
• 2020

In fine-tuning-based transfer learning, the size of the dataset may affect learning accuracy. When a dataset scale is small, fine-tuning-based transfer-learning methods use high computing costs, similar to a large-scale dataset. We propose a mid-level feature extractor that retrains only the mid-level convolutional layers, resulting in increased efficiency and reduced computing costs. This mid-level feature extractor is likely to provide an effective alternative in training a small-scale medical image dataset. The performance of the mid-level feature extractor is compared with the performance of low- and high-level feature extractors, as well as the fine-tuning method. First, the mid-level feature extractor takes a shorter time to converge than other methods do. Second, it shows good accuracy in validation loss evaluation. Third, it obtains an area under the ROC curve (AUC) of 0.87 in an untrained test dataset that is very different from the training dataset. Fourth, it extracts more clear feature maps about shape and part of the chest in the X-ray than fine-tuning method.

• ETRI Journal
• /
• v.34 no.6
• /
• pp.847-857
• /
• 2012

Intrusion detection systems (IDSs) have an important effect on system defense and security. Recently, most IDS methods have used transformed features, selected features, or original features. Both feature transformation and feature selection have their advantages. Neighborhood component analysis feature transformation and genetic feature selection (NCAGAFS) is proposed in this research. NCAGAFS is based on soft computing and data mining and uses the advantages of both transformation and selection. This method transforms features via neighborhood component analysis and chooses the best features with a classifier based on a genetic feature selection method. This novel approach is verified using the KDD Cup99 dataset, demonstrating higher performances than other well-known methods under various classifiers have demonstrated.

• Proceedings of the Korea Inteligent Information System Society Conference
• /
• 2001.01a
• /
• pp.421-427
• /
• 2001

In this paper, a method for selection of the optimal feature vectors is proposed for the classification of closed 2D shapes using the bispectrum of a contour sequence. The bispectrum based on third order cumulants is applied to the contour sequences of the images to extract feature vectors for each planar image. These bispectral feature vectors, which are invariant to shape translation, rotation and scale transformation, can be used to represent two-dimensional planar images, but there is no certain criterion on the selection of the feature vectors for optimal classification of closed 2D images. In this paper, a new method for selecting the optimal bispectral feature vectors based on the variances of the feature vectors. The experimental results are presented using eight different shapes of aircraft images, the feature vectors of the bispectrum from five to fifteen and an weighted mean fuzzy classifier.

• Journal of Korea Multimedia Society
• /
• v.18 no.7
• /
• pp.834-842
• /
• 2015

We propose a method to construct composite feature vector based on discriminant analysis for face recognition. For this, we first extract the holistic- and local-features from whole face images and local images, which consist of the discriminant pixels, by using a discriminant feature extraction method. In order to utilize both advantages of holistic- and local-features, we evaluate the amount of the discriminative information in each feature and then construct a composite feature vector with only the features that contain a large amount of discriminative information. The experimental results for the FERET, CMU-PIE and Yale B databases show that the proposed composite feature vector has improvement of face recognition performance.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.12
• /
• pp.6028-6042
• /
• 2019

Since Sparse Nonnegative Matrix Factorization (SNMF) method can control the sparsity of the decomposed matrix, and then it can be adopted to control the sparsity of facial feature extraction and recognition. In order to improve the accuracy of SNMF method for facial feature recognition, new additive iterative rules based on the improved iterative step sizes are proposed to improve the SNMF method, and then the traditional multiplicative iterative rules of SNMF are transformed to additive iterative rules. Meanwhile, to further increase the sparsity of the basis matrix decomposed by the improved SNMF method, a threshold-sparse constraint is adopted to make the basis matrix to a zero-one matrix, which can further improve the accuracy of facial feature recognition. The improved SNMF method based on the additive iterative rules and threshold-sparse constraint is abbreviated as ASNMF, which is adopted to recognize the ORL and CK+ facial datasets, and achieved recognition rate of 96% and 100%, respectively. Meanwhile, from the results of the contrast experiments, it can be found that the recognition rate achieved by the ASNMF method is obviously higher than the basic NMF, traditional SNMF, convex nonnegative matrix factorization (CNMF) and Deep NMF.

• Proceedings of the IEEK Conference
• /
• 2000.11d
• /
• pp.13-16
• /
• 2000

This paper presents a robust method to combine a collection of images with small fields of view to obtain an image with a large field of view. In the previous works, there are two main areas which one is a cross correlation-based method and the other is a feature-based method. The former is based on motion estimation from video sequences. so there are a problem on rotating a camera about optical axis. In the latter method, it is difficult to match correspondence feature points correctly.'re find correct correspondences, we proposed the geometrical feature model and correspondence filters and the Gaussian distribution weight function to blend the images smoothly. The experiments show that our method is robust and effective.