• Journal of Electrical Engineering and Technology
• /
• v.10 no.4
• /
• pp.1851-1863
• /
• 2015

– In the paper, to handle the problem of traditional bag-of-features model ignoring the spatial relationship of local features in human action recognition, we proposed a Multiscale Spatial Position Coding under Locality Constraint method. Specifically, to describe this spatial relationship, we proposed a mixed feature combining motion feature and multi-spatial-scale configuration. To utilize temporal information between features, sub spatial-temporal-volumes are built. Next, the pooled features of sub-STVs are obtained via max-pooling method. In classification stage, the Locality-Constrained Group Sparse Representation is adopted to utilize the intrinsic group information of the sub-STV features. The experimental results on the KTH, Weizmann, and UCF sports datasets show that our action recognition system outperforms the classical local ST feature-based recognition systems published recently.

• International Journal of Contents
• /
• v.8 no.4
• /
• pp.56-63
• /
• 2012

In recent decades, analyzing the activities of human brain achieved some accomplishments by using the functional Magnetic Resonance Imaging (fMRI) technique. fMRI data provide a sequence of three-dimensional images related to human brain's activity which can be used to detect instantaneous cognitive states by applying machine learning methods. In this paper, we propose a new approach for distinguishing human's cognitive states such as "observing a picture" versus "reading a sentence" and "reading an affirmative sentence" versus "reading a negative sentence". Since fMRI data are high dimensional (about 100,000 features in each sample), extremely sparse and noisy, feature selection is a very important step for increasing classification accuracy and reducing processing time. We used the Fisher Discriminant Ratio to select the most powerful discriminative features from some Regions of Interest (ROIs). The experimental results showed that our approach achieved the best performance compared to other feature extraction methods with the average accuracy approximately 95.83% for the first study and 99.5% for the second study.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.3
• /
• pp.1403-1417
• /
• 2019

With the rapid growth of multimedia data, research on cross-media feature learning has significance in many applications, such as multimedia search and recommendation. Existing methods are sensitive to noise and edge information in multimedia data. In this paper, we propose a semi-supervised method for cross-media feature learning by means of $L_{2,q}$ norm to improve the performance of cross-media retrieval, which is more robust and efficient than the previous ones. In our method, noise and edge information have less effect on the results of cross-media retrieval and the dynamic patch information of multimedia data is employed to increase the accuracy of cross-media retrieval. Our method can reduce the interference of noise and edge information and achieve fast convergence. Extensive experiments on the XMedia dataset illustrate that our method has better performance than the state-of-the-art methods.

• Journal of Power Electronics
• /
• v.19 no.2
• /
• pp.443-453
• /
• 2019

As the core component of transmission systems, converters are very prone to failure. To improve the accuracy of fault diagnosis for wind power converters, a fault feature extraction method combined with a wavelet transform and compressed sensing theory is proposed. In addition, an improved AdaBoost-SVM is used to diagnose wind power converters. The three-phase output current signal is selected as the research object and is processed by the wavelet transform to reduce the signal noise. The wavelet approximation coefficients are dimensionality reduced to obtain measurement signals based on the theory of compressive sensing. A sparse vector is obtained by the orthogonal matching pursuit algorithm, and then the fault feature vector is extracted. The fault feature vectors are input to the improved AdaBoost-SVM classifier to realize fault diagnosis. Simulation results show that this method can effectively realize the fault diagnosis of the power transistors in converters and improve the precision of fault diagnosis.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.37 no.7C
• /
• pp.554-564
• /
• 2012

The vehicle recognition consists of two steps; the vehicle region detection step and the vehicle identification step based on the feature extracted from the detected region. Features using linear transformations have the effect of dimension reduction as well as represent statistical characteristics, and show the robustness in translation and rotation of objects. Among the linear transformations, the NMF(Non-negative Matrix Factorization) is one of part-based representation. Therefore, we can extract NMF features with sparsity and improve the vehicle recognition rate by the representation of local features of a car as a basis vector. In this paper, we propose a feature extraction using NMF suitable for the vehicle recognition, and verify the recognition rate with it. Also, we compared the vehicle recognition rate for the occluded area using the SNMF(sparse NMF) which has basis vectors with constraint and LVQ2 neural network. We showed that the feature through the proposed NMF is robust in the urban scene where occlusions are frequently occur.

• Journal of Information Processing Systems
• /
• v.16 no.1
• /
• pp.6-29
• /
• 2020

Biometrics identification using multiple modalities has attracted the attention of many researchers as it produces more robust and trustworthy results than single modality biometrics. In this paper, we present a novel multimodal recognition system that trains a deep learning network to automatically learn features after extracting multiple biometric modalities from a single data source, i.e., facial video clips. Utilizing different modalities, i.e., left ear, left profile face, frontal face, right profile face, and right ear, present in the facial video clips, we train supervised denoising auto-encoders to automatically extract robust and non-redundant features. The automatically learned features are then used to train modality specific sparse classifiers to perform the multimodal recognition. Moreover, the proposed technique has proven robust when some of the above modalities were missing during the testing. The proposed system has three main components that are responsible for detection, which consists of modality specific detectors to automatically detect images of different modalities present in facial video clips; feature selection, which uses supervised denoising sparse auto-encoders network to capture discriminative representations that are robust to the illumination and pose variations; and classification, which consists of a set of modality specific sparse representation classifiers for unimodal recognition, followed by score level fusion of the recognition results of the available modalities. Experiments conducted on the constrained facial video dataset (WVU) and the unconstrained facial video dataset (HONDA/UCSD), resulted in a 99.17% and 97.14% Rank-1 recognition rates, respectively. The multimodal recognition accuracy demonstrates the superiority and robustness of the proposed approach irrespective of the illumination, non-planar movement, and pose variations present in the video clips even in the situation of missing modalities.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.1
• /
• pp.302-320
• /
• 2017

In this paper, an innovative robust feature detection and matching strategy for visual odometry based on stereo image sequence is proposed. First, a sparse multiscale 2D local invariant feature detection and description algorithm AKAZE is adopted to extract the interest points. A robust feature matching strategy is introduced to match AKAZE descriptors. In order to remove the outliers which are mismatched features or on dynamic objects, an improved random sample consensus outlier rejection scheme is presented. Thus the proposed method can be applied to dynamic environment. Then, geometric constraints are incorporated into the motion estimation without time-consuming 3-dimensional scene reconstruction. Last, an iterated sigma point Kalman Filter is adopted to refine the motion results. The presented ego-motion scheme is applied to benchmark datasets and compared with state-of-the-art approaches with data captured on campus in a considerably cluttered environment, where the superiorities are proved.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.7
• /
• pp.1-7
• /
• 2021

In this paper, we propose a feature selection technique for multi-label classification. Many existing feature selection techniques have selected features by calculating the relation between features and labels such as a mutual information scale. However, since the mutual information measure requires a joint probability, it is difficult to calculate the joint probability from an actual premise feature set. Therefore, it has the disadvantage that only a few features can be calculated and only local optimization is possible. Away from this regional optimization problem, we propose a feature selection technique that constructs a low-rank space in the entire given feature space and selects features with sparsity. To this end, we designed a regression-based objective function using Nuclear norm, and proposed an algorithm of gradient descent method to solve the optimization problem of this objective function. Based on the results of multi-label classification experiments on four data and three multi-label classification performance, the proposed methodology showed better performance than the existing feature selection technique. In addition, it was showed by experimental results that the performance change is insensitive even to the parameter value change of the proposed objective function.

• Journal of Information Processing Systems
• /
• v.4 no.2
• /
• pp.77-86
• /
• 2008

This research proposes a new neural network for text categorization which uses alternative representations of documents to numerical vectors. Since the proposed neural network is intended originally only for text categorization, it is called NTC (Neural Text Categorizer) in this research. Numerical vectors representing documents for tasks of text mining have inherently two main problems: huge dimensionality and sparse distribution. Although many various feature selection methods are developed to address the first problem, the reduced dimension remains still large. If the dimension is reduced excessively by a feature selection method, robustness of text categorization is degraded. Even if SVM (Support Vector Machine) is tolerable to huge dimensionality, it is not so to the second problem. The goal of this research is to address the two problems at same time by proposing a new representation of documents and a new neural network using the representation for its input vector.

• KIPS Transactions on Software and Data Engineering
• /
• v.4 no.4
• /
• pp.187-194
• /
• 2015

In this paper, we present an effective visual odometry estimation system to track the real-time pose of a camera moving in 3D space. In order to meet the real-time requirement as well as to make full use of rich information from color and depth images, our system adopts a feature-based sparse odometry estimation method. After matching features extracted from across image frames, it repeats both the additional inlier set refinement and the motion refinement to get more accurate estimate of camera odometry. Moreover, even when the remaining inlier set is not sufficient, our system computes the final odometry estimate in proportion to the size of the inlier set, which improves the tracking success rate greatly. Through experiments with TUM benchmark datasets and implementation of the 3D scene reconstruction application, we confirmed the high performance of the proposed visual odometry estimation method.