• Journal of Korea Multimedia Society
• /
• v.12 no.10
• /
• pp.1418-1426
• /
• 2009

There has been a lot of interest in an effective method for background subtraction in an effort to separate foreground objects from a predefined background image. Promising results on background subtraction using statistical methods have recently been reported are robust enough to operate in dynamic environments, but generally require very large computational resources and still have difficulty in obtaining clear segmentation of objects. We use a simple running-average method to model a gradually changing background, instead of using a complicated statistical technique. We employ a single global threshold vector, optimized by a genetic algorithm, instead of pixel-by-pixel thresholds. A new fitness function is defined and trained to evaluate segmentation result. The system has been implemented on a PC with a webcam, and experimental results on real images show that the new method outperforms an existing method based on a mixture of Gaussian.

• The KIPS Transactions:PartD
• /
• v.11D no.5
• /
• pp.1177-1188
• /
• 2004

Multi-agent systems have been developed for supporting intelligent collaboration of distributed and independent software entities and are be-ing widely used for various applications. For the collaboration among agents, conversation policies (or interaction protocols) mutually agreed by agents are used. In today's dynamic electronic market environment, there can be frequent changes in conversation policies induced by the changes in transaction methods in the market, and thus, the importance of ad hoc conversation policies is increasing. In existing agent platforms, they allow the use of only several standard or fixed conversation policies, which requires inevitable re implementation for ad hoc conversation policies and leads to inefficiency and intricacy. This paper designs an agent platform that supports ad hoc conversation policies and presents the prototype implementation. The suggested system includes an exchangeable and interpretable conversation policy model, a meta conversation procedure for exchanging new conversation policies, and a mechanism for performing actual transactions with exchanged conversation policies in run time in an adaptive way.

• Proceedings of the Korea Society for Simulation Conference
• /
• 2001.10a
• /
• pp.321-324
• /
• 2001

The robot soccer simulation game is a dynamic multi-agent environment. In this paper we suggest a new reinforcement learning approach to each agent's dynamic positioning in such dynamic environment. Reinforcement Beaming is the machine learning in which an agent learns from indirect, delayed reward an optimal policy to choose sequences of actions that produce the greatest cumulative reward. Therefore the reinforcement loaming is different from supervised teaming in the sense that there is no presentation of input-output pairs as training examples. Furthermore, model-free reinforcement loaming algorithms like Q-learning do not require defining or loaming any models of the surrounding environment. Nevertheless it can learn the optimal policy if the agent can visit every state-action pair infinitely. However, the biggest problem of monolithic reinforcement learning is that its straightforward applications do not successfully scale up to more complex environments due to the intractable large space of states. In order to address this problem, we suggest Adaptive Mediation-based Modular Q-Learning(AMMQL) as an improvement of the existing Modular Q-Learning(MQL). While simple modular Q-learning combines the results from each learning module in a fixed way, AMMQL combines them in a more flexible way by assigning different weight to each module according to its contribution to rewards. Therefore in addition to resolving the problem of large state space effectively, AMMQL can show higher adaptability to environmental changes than pure MQL. This paper introduces the concept of AMMQL and presents details of its application into dynamic positioning of robot soccer agents.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.39 no.1
• /
• pp.51-65
• /
• 2002

This paper presents a design method of the neural network-based controller using an indirect adaptive control method to deal with an intelligent control for chaotic nonlinear systems. The proposed control method includes the identification and control Process for chaotic nonlinear systems. The identification process for chaotic nonlinear systems is an off-line process which utilizes the serial-parallel structure of multilayer neural networks and simple state space neural networks. The control process is an on-line process which uses the trained neural networks as the system model. An error back-propagation method was used for training of identification and control for chaotic nonlinear systems. The performance of the proposed neural network controller was evaluated by application to the Duffing equation and the Lorenz equation, and the proposed controller was compared with other neural network-based controllers by computer simulations.

• The Journal of the Acoustical Society of Korea
• /
• v.22 no.2
• /
• pp.135-144
• /
• 2003

A method for the automatic segmentation of speech signals is described. The method is dedicated to the construction of a large database for a Text-To-Speech (TTS) synthesis system. The main issue of the work involves the refinement of an initial estimation of phone boundaries which are provided by an alignment, based on a Hidden Market Model(HMM). Multi-layer perceptron (MLP) was used as a phone boundary detector. To increase the performance of segmentation, a technique which individually trains an MLP according to phonetic transition is proposed. The optimum partitioning of the entire phonetic transition space is constructed from the standpoint of minimizing the overall deviation from hand labelling positions. With single speaker stimuli, the experimental results showed that more than 95% of all phone boundaries have a boundary deviation from the reference position smaller than 20 ms, and the refinement of the boundaries reduces the root mean square error by about 25%.

• The Journal of the Acoustical Society of Korea
• /
• v.11 no.2
• /
• pp.59-67
• /
• 1992

Recently, the problem which actively control the unwanted noise propagated from the technical structure by the generated secondary sound has become considerable topic from the environmental preservation point of view. In most of these studies, active noise control deals with a plane wave propagation at low frequency using adaptive filtering techniques. On the other hand, in real acoustic systems are mostly short due to the limitation of geometric configuration. In this case, the acoustic properties such as reflections and resonances inside the acoustic system should be considered. In this paper, the acoustic modeling method for short length duct was introduced using the transfer matrix method, and the active noise control problem was investigated with \implementation of FIR filter for the transfer function of control system derived from this modeling method. The identification methods for the acoustic model of actual control system was proposed by numerical computation technique based on the estimation of optimal FIR filter coefficients. The acceptable attenuation on the real acoustic system and stability of the controller are predicted in this computational simulation.

• Journal of Digital Convergence
• /
• v.15 no.6
• /
• pp.229-236
• /
• 2017

An efficient frequent pattern algorithm is essential for mining association rules as well as many other mining tasks for convergence with its application spread over a very broad spectrum. Models for mining pattern have been proposed using a FP-tree for storing compressed information about frequent patterns. In this paper, we propose a centroid frequent pattern growth algorithm which we called "CAWFP-Growth" that enhances he FP-Growth algorithm by making the center of weights and frequencies for the itemsets. Because the conventional constraint of maximum weighted support is not necessary to maintain the downward closure property, it is more likely to reduce the search time and the information loss of the frequent patterns. The experimental results show that the proposed algorithm achieves better performance than other algorithms without scarifying the accuracy and increasing the processing time via the centroid of the items. The MapReduce framework model is provided to handle large amounts of data via a pseudo-distributed computing environment. In addition, the modeling of the proposed algorithm is required in the fully distributed mode.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.9
• /
• pp.208-214
• /
• 2014

Vehicle dynamic states used in various advanced driving safety systems are influenced by road geometry. Among the road geometry information, the vehicle pitch angle influenced by road slope and acceleration-deceleration is essential parameter used in pose estimation including the navigation system, advanced adaptive cruise control and others on sag road. Although the road slope data is essential parameter, the method measuring the parameter is not commercialized. The digital map including the road geometry data and high-precision DGPS system such as DGPS(Differential Global Positioning System) based RTK(Real-Time Kinematics) are used unusually. In this paper, low-cost cascade extended Kalman filter(CEKF) based road slope estimation method is proposed. It use cascade two EKFs. The EKFs use several measured vehicle states such as yaw rate, longitudinal acceleration, lateral acceleration and wheel speed of the rear tires and 3 D.O.F(Degree Of Freedom) vehicle dynamics model. The performance of proposed estimation algorithm is evaluated by simulation based on Carsim dynamics tool and T-car based experiment.

• Korean Journal of Social Welfare
• /
• v.66 no.3
• /
• pp.5-28
• /
• 2014

School absenteeism, one of the early warning signs of behavioral problems, has been known to be a complex and heterogeneous problem. However, much of the research assumes that school absentee adolescents comprise a homogeneous group. This study explored the heterogeneity of school absentee adolescents, based on school adaptation, to provide a more nuanced understanding of school absenteeism and examined predictive and risk factors associated with each typology of school absentee adolescents. Latent profile analysis was conducted using sample 477 middle school students who were reported absent in the previous year from the 3rd wave of Korean Children and Youth Panel Study. Multinomial logistic regression and ANOVA was also employed to examine the effects of predictive variables. As a result, three profiles were identified: low, middle, and high adaptive group. Group membership was found to be associated deferentially with gender, mental health, parenting neglect, delinquent friends, and delinquent behaviors. These findings propose more specific and targeted interventions designed to meet the needs and risk factors associated with the different typologies of school absentee adolescents.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.16 no.2
• /
• pp.237-245
• /
• 2012

This paper describes an acoustic echo canceller with double-talk by the blind signal separation. The acoustic echo canceller is deteriorated or diverged in the double-talk period. So we use the blind signal separation to estimate the near-end speech signal and to eliminate the estimated signal from the residual signal. The blind signal separation extracts the near-end signal with dual microphones by the iterative computations using the 2nd order statistical character. Because the mixture model of blind signal separation is multi-channel in the closed reverberation environment, we used the copied coefficients of echo canceller without computing the separation coefficients. By this method, the acoustic echo canceller operates irrespective of double-talking. We verified performances of the proposed acoustic echo canceller by simulations. The results show that the acoustic echo canceller with this algorithm detects the double-talk periods thoroughly, and then operates stably in the normal state without the divergence of coefficients after ending the double-talking. And it shows the ERLE of averagely 20dB higher than the normal LMS algorithm.