• Journal of information and communication convergence engineering
• /
• v.15 no.2
• /
• pp.73-78
• /
• 2017

A general linearly constrained broadband adaptive array is examined in the eigenvector space with respect to the optimal weight vector and the adaptive algorithm. The optimal weight vector and the general adaptive algorithm in the eigenvector space are obtained by eigenvector matrix transformation. Their operations are shown to be the same as in the standard coordinate system except for the relevant transformed vectors and matrices. The nulling performance of the general linearly constrained broadband adaptive array depends on the gain factor such that the constraint plane is shifted perpendicularly to the origin by an increase in the gain factor. The general linearly constrained broadband adaptive array is observed to perform better than a conventional linearly constrained adaptive array in a coherent signal environment, while the former performs similarly to the latter in a non-coherent signal environment.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2002.05a
• /
• pp.509-513
• /
• 2002

This paper presents a semi-blind Minimum Mean Square Error (MMSE) beamforming adaptive algorithm used far OFDM/CDMA combined system. The proposed algorithm exploits the transmitting pilot signal in the initial period of the transmission to update the weight vector. Then it applies the blind adaptive period to update the weight vector, in which the pilot signal is no longer used. The derivation of the algorithm based on the Mean Square Error (MSE) criterion is also presented. Computer simulation is carried out to verify the performance of the proposed approach.

• Proceedings of the KIEE Conference
• /
• 2007.10a
• /
• pp.407-408
• /
• 2007

In this paper, we present a new method for classifying object using LVQ (Learning Vector Quantization) with flexible output layer. The proposed LVQ is a supervised learning method that dynamically generates output neurons and initializes automatically the weight vectors from training patterns. If the classes of the nearest output neuron is different from the class of the training pattern, a new output neuron is created and the given training pattern is used to initialize the weight vector of the created neuron. The proposed method is significantly different from the previous competitive learning algorithms in the point that the output neurons are dynamically generated during the learning process.

• Annual Conference of KIPS
• /
• 2004.05a
• /
• pp.1573-1576
• /
• 2004

멀티미디어 압축 및 이동 통신 기술의 발전으로 다양한 형태의 멀티미디어 서비스가 이슈화되고 있다. 비디오를 전송하기 위해서는 많은 대역폭을 필요로 하지만, 모든 네트워크가 높은 수준의 대역 및 처리 능력을 가지는 것은 아니다. 이질적인 네트워크간의 멀티미디어를 서비스하기 위해서는 네트워크 상황 또는 수신자의 처리 능력에 맞도록 재 부호화해야 하지만 그 처리비용이 높다. 트랜스코딩 기법 중에서 시간당 요구된 프레임의 개수를 조절하면 제거된 프레임의 움직임 벡터를 재 사용하여 비트율을 감소시킬 수 있다. 본 논문에서는 기존의 기법보다 향상된 움직임과 중첩 영역의 정보를 적용한 WBVC(Weight-Based Vector Composition) 기법을 제안한다. 실험을 통한 기존의 기법과의 비교 분석 결과, 비슷한 계산 복잡도에서 제안한 WBVC 기법이 높은 성능을 보였다.

• Annual Conference on Human and Language Technology
• /
• 2019.10a
• /
• pp.326-329
• /
• 2019

기계독해를 실용화하기 위해 단락을 검색하는 검색 모델은 최근 기계독해 모델이 우수한 성능을 보임에 따라 그 필요성이 더 부각되고 있다. 그러나 기존 검색 모델은 질의와 단락의 어휘 일치도나 유사도만을 계산하므로, 기계독해에 필요한 질의 어휘의 문맥에 해당하는 단락 검색을 하지 못하는 문제가 있다. 본 논문에서는 이러한 문제를 해결하기 위해 Word2vec의 입력 단어열의 벡터에 해당하는 IN Weight Matrix와 출력 단어열의 벡터에 해당하는 OUT Weight Matrix를 사용한 단락 검색 모델을 제안한다. 제안 방법은 기존 검색 모델에 비해 정확도를 측정하는 Precision@k에서 좋은 성능을 보였다.

• 제어로봇시스템학회:학술대회논문집
• /
• 2002.10a
• /
• pp.111.1-111
• /
• 2002

System configuration 1. First is the image acquisition part 2. Second part is for creating the vector image and for processing the obtained facial image. This part is for finding the facial area from the skin color. To do this, we can first find the skin color area with the highest weight from eigenface that consists of eigenvector. And then, we can create the vector image of eigenface from the obtained facial area. 3. Third is recognition module portion.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.11 no.7
• /
• pp.675-679
• /
• 2001

In this paper, we present a new competitive learning algorithm called Dynamic Competitive Learning (DCL). DCL is a supervised learning method that dynamically generates output neurons and initializes automatically the weight vectors from training patterns. It introduces a new parameter called LOG (Limit of Grade) to decide whether an output neuron is created or not. If the class of at least one among the LOG number of nearest output neurons is the same as the class of the present training pattern, then DCL adjusts the weight vector associated with the output neuron to learn the pattern. If the classes of all the nearest output neurons are different from the class of the training pattern, a new output neuron is created and the given training pattern is used to initialize the weight vector of the created neuron. The proposed method is significantly different from the previous competitive learning algorithms in the point that the selected neuron for learning is not limited only to the winner and the output neurons are dynamically generated during the learning process. In addition, the proposed algorithm has a small number of parameters, which are easy to be determined and applied to real-world problems. Experimental results for pattern recognition of remote sensing data and handwritten numeral data indicate the superiority of DCL in comparison to the conventional competitive learning methods.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.28 no.10C
• /
• pp.984-994
• /
• 2003

This paper proposes a novel blind adaptive algorithm for computing the weight vector of an antenna array system. The new technique utilizes a Generalized On-Off algorithm to obtain the weight vector maximizing the SINR(Signal to Interference plus Noise Ratio) of the received signal. It is observed that the proposed algorithm generates a suboptimal weight vector with a linear computational load(O(6N+8)). From the various simulations, it is confirmed that, when the signal environment becomes adverse, e.g., low Processing Gain, and/or wide angular spread. the proposed algorithm outperforms the conventional one in terms of the communication capacity by about 3 times. Applying the proposed algorithm to satellite tracking systems as well as IS2000 1X mobile communication system, we have found that both communication capacity and communication quality are significantly improved.

• The Korean Journal of Applied Statistics
• /
• v.16 no.2
• /
• pp.261-271
• /
• 2003

Support Vector Machines (SVMs) provide a powerful performance of the learning system. Generally, SVMs tend to make overfitting. For the purpose of overcoming this difficulty, the definition of soft margin has been introduced. In this case, it causes another difficulty to decide the weight for slack variables reflecting soft margin classifiers. Especially, the error of soft margin algorithm can be bounded by a target margin and some norms of the slack vector. In this paper, we formulate a new soft margin algorithm considering the bound of corruption by noise in data directly. Additionally, through a numerical example, we compare the proposed method with a conventional soft margin algorithm.

• Journal of Broadcast Engineering
• /
• v.3 no.2
• /
• pp.118-126
• /
• 1998

In order to fedice the computational complexity of the motion estimation for the spatio-temporal prediction of the higher layer, two estimation method are proposed. In the first one, the motion vector of the higher layer is estimated within the small search range by using the previously estimated motion vector in the lower layer as an innitial vector. Inthe second one, the notion vector is estimated by the spatio-temporally weighted search, which is combined with the previously estimated motion vector of the lower layer and the weight for spatial prediction. Simulation results show that the proposed methods give the smaller computational complexity without the degradation of the coding efficiency than the conventinal one.