• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.1
• /
• pp.368-391
• /
• 2018

Face recognition (FR) with a single sample per person (SSPP) is common in real-world face recognition applications. In this scenario, it is hard to predict intra-class variations of query samples by gallery samples due to the lack of sufficient training samples. Inspired by the fact that similar faces have similar intra-class variations, we propose a virtual sample generating algorithm called k nearest neighbors based virtual sample generating (kNNVSG) to enrich intra-class variation information for training samples. Furthermore, in order to use the intra-class variation information of the virtual samples generated by kNNVSG algorithm, we propose image set based multimanifold discriminant learning (ISMMDL) algorithm. For ISMMDL algorithm, it learns a projection matrix for each manifold modeled by the local patches of the images of each class, which aims to minimize the margins of intra-manifold and maximize the margins of inter-manifold simultaneously in low-dimensional feature space. Finally, by comprehensively using kNNVSG and ISMMDL algorithms, we propose k nearest neighbor virtual image set based multimanifold discriminant learning (kNNMMDL) approach for single sample face recognition (SSFR) tasks. Experimental results on AR, Multi-PIE and LFW face datasets demonstrate that our approach has promising abilities for SSFR with expression, illumination and disguise variations.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.3
• /
• pp.1121-1141
• /
• 2020

We propose a robust visual object tracking algorithm fusing a convolutional neural network tracker trained offline from a large number of video repositories and a color histogram based tracker to track objects for mixing immersive audio. Our algorithm addresses the problem of occlusion and large movements of the CNN based GOTURN generic object tracker. The key idea is the offline training of a binary classifier with the color histogram similarity values estimated via both trackers used in this method to opt appropriate tracker for target tracking and update both trackers with the predicted bounding box position of the target to continue tracking. Furthermore, a histogram similarity constraint is applied before updating the trackers to maximize the tracking accuracy. Finally, we compute the depth(z) of the target object by one of the prominent unsupervised monocular depth estimation algorithms to ensure the necessary 3D position of the tracked object to mix the immersive audio into that object. Our proposed algorithm demonstrates about 2% improved accuracy over the outperforming GOTURN algorithm in the existing VOT2014 tracking benchmark. Additionally, our tracker also works well to track multiple objects utilizing the concept of single object tracker but no demonstrations on any MOT benchmark.

• Proceedings of the Korean Society for Bioinformatics Conference
• /
• 2000.11a
• /
• pp.59-60
• /
• 2000

The past few years have seen a dramatic increase in gene expression data on the basis of DNA microarrays or DNA chips. Going beyond a generic view on the genome, microarray data are able to distinguish between gene populations in different tissues of the same organism and in different states of cells belonging to the same tissue. This affords a cell-wide view of the metabolic and regulatory processes under different conditions, building an effective basis for new diagnoses and therapies of diseases. In this talk we present machine learning techniques for effective mining of DNA microarray data. A brief introduction to the research field of machine learning from the computer science and artificial intelligence point of view is followed by a review of recently-developed learning algorithms applied to the analysis of DNA chip gene expression data. Emphasis is put on graphical models, such as Bayesian networks, latent variable models, and generative topographic mapping. Finally, we report on our own results of applying these learning methods to two important problems: the identification of cell cycle-regulated genes and the discovery of cancer classes by gene expression monitoring. The data sets are provided by the competition CAMDA-2000, the Critical Assessment of Techniques for Microarray Data Mining.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2002.05a
• /
• pp.559-564
• /
• 2002

Cross-talk cancellation, inverse filter design or deconvolution in a generic term, is a vital process for a virtual sound realization in the stereo sound reproduction system. Most, if not all, of the design algorithms available for the inverse filter are based on a linearized model of the real physical plant. The result of such a plant-based design method, which may be referred to here as the indirect method, is biased due to both modelling and inversion errors. This paper presents a novel direct cross-talk cancellation method that may be free from the inversion error. The direct method can directly models the inverse filter by a suitable rearrangement of the input and output ports of the original plant so that no inversion is required here. Advantages are discussed with various experiments in an anechoic chamber using a PC soundcard. Binaural reproduction tests conducted showed that the conventional indirect method yields about 8 % reproduction performance error on both ear positions, whereas the direct method offers about 3 %.

• Journal of the Korean Society for Aviation and Aeronautics
• /
• v.16 no.2
• /
• pp.1-11
• /
• 2008

Increasing demand for improved reliability and survivability of mission-critical systems is driving the development of health monitoring and Automated Contingency Management (ACM) systems. An ACM system is expected to adapt autonomously to fault conditions with the goal of still achieving mission objectives by allowing some degradation in system performance within permissible limits. ACM performance depends on supporting technologies like sensors and anomaly detection, diagnostic/prognostic and reasoning algorithms. This paper presents the development of a generic prototype test bench software framework for developing and validating ACM systems for advanced propulsion systems called the Propulsion ACM (PACM) Test Bench. The architecture has been implemented for a Monopropellant Propulsion System (MPS) to demonstrate the validity of the approach. A Simulink model of the MPS has been developed along with a fault injection module. It has been shown that the ACM system is capable of mitigating the failures by searching for an optimal strategy. Furthermore, the concepts of Validation and Verification (V&V) of such systems are introduced with relevant examples.

• Proceedings of the KIEE Conference
• /
• 2001.07a
• /
• pp.6-9
• /
• 2001

This paper presents a new method which applies a genetic algorithm(GA) for determining which sectionalizing switch to operate in order to solve the distribution system loss minimization re-configuration problem. The distribution system loss minimization re-configuration problem is in essence a 0-1 planning problem which means that for typical system scales the number of combinations requiring searches becomes extremely large. In order to deal with this problem, a new approach which applies a GA was presented. Briefly, GA are a type of random number search method, however, they incorporate a multi-point search feature. Further, every point is not is not separately and respectively renewed, therefore, if parallel processing is applied, we can expect a fast solution algorithm to result.

• Proceedings of the KIEE Conference
• /
• 2002.07d
• /
• pp.2821-2823
• /
• 2002

In this paper we propose the Multi-FNN (Fuzzy-Neural Networks) for optimal identification modeling of complex system. The proposed Multi-FNNs is based on a concept of FNNs and exploit linear inference being treated as generic inference mechanisms. In the networks learning, backpropagation(BP) algorithm of neural networks is used to updata the parameters of the network in order to control of nonlinear process with complexity and uncertainty of data, proposed model use a HCM(Hard C-Means)clustering algorithm which carry out the input-output dat a preprocessing function and Genetic Algorithm which carry out optimization of model The HCM clustering method is utilized to determine the structure of Multi-FNNs. The parameters of Multi-FNN model such as apexes of membership function, learning rates, and momentum coefficients are adjusted using genetic algorithms. An aggregate performance index with a weighting factor is proposed in order to achieve a sound balance between approximation and generalization abilities of the model. NOx emission process data of gas turbine power plant is simulated in order to confirm the efficiency and feasibility of the proposed approach in this paper.

• ETRI Journal
• /
• v.28 no.5
• /
• pp.593-606
• /
• 2006

Thanks to the great possibilities of providing different types of telecommunication traffic to a large geographical area, satellite networks are expected to be an essential component of the next-generation internet. As a result, issues concerning the designing and testing of efficient connection-admission-control (CAC) strategies in order to increase the quality of service (QoS) for multimedia traffic sources, are attractive and at the cutting edge of research. This paper investigates the potential strengths of a generic digital-video-broadcasting return-channel-via-satellite (DVB-RCS) system architecture, proposing a new CAC algorithm with the aim of efficiently managing real-time multimedia video sources, both with constant and high variable data rate transmission; moreover, the proposed admission strategy is compared with a well-known iterative CAC mainly designed for the managing of real-time bursty traffic sources in order to demonstrate that the new algorithm is also well suited for those traffic sources. Performance analysis shows that, both algorithms guarantee the agreed QoS to real-time bursty connections that are more sensitive to delay jitter; however, our proposed algorithm can also manage interactive real-time multimedia traffic sources in high load and mixed traffic conditions.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2002.04a
• /
• pp.647-650
• /
• 2002

The human faces do not have distinct features unlike other general objects. In general the features of eyes, nose and mouth which are first recognized when human being see the face are defined. These features have different characteristics depending on different human face. In this paper, We propose a face recognition algorithm using the hidden Markov model(HMM). In the preprocessing stage, we find edges of a face using the locally adaptive threshold scheme and extract features based on generic knowledge of a face, then construct a database with extracted features. In training stage, we generate HMM parameters for each person by using the forward-backward algorithm. In the recognition stage, we apply probability values calculated by the HMM to input data. Then the input face is recognized by the euclidean distance of face feature vector and the cross-correlation between the input image and the database image. Computer simulation shows that the proposed HMM algorithm gives higher recognition rate compared with conventional face recognition algorithms.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.28 no.1B
• /
• pp.27-32
• /
• 2003

This paper presents the Genetic Algorithm based Task Scheduling (GATS) method for the scheduling of programs with diverse embedded parallelism types in Distributed Parallel Systems, which consist of a set of loosely coupled parallel and vector machines connected via high speed networks The distributed parallel processing tries to solve computationally intensive problems that have several types of parallelism, on a suite of high performance and parallel machines in a manner that best utilizes the capabilities of each machine. When scheduling in distributed parallel systems, the matching of the parallelism characteristics between tasks and parallel machines rather than load balancing should be carefully handled with the minimization of communication cost in order to obtain more speedup. This paper proposes the based initialization methods for an initial population and the knowledge-based mutation methods to accommodate the parallelism type matching in genetic algorithms.