• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2006.11a
• /
• pp.297-300
• /
• 2006

본 논문에서는 비선형 기동표적의 추적에 대한 새로운 접근 방식을 소개한다. 이 논문에서는 표적의 가속도를 시변 변수인 표적의 추가적인 잡음으로 두고 각각의 가속도 간격의 정도에 따라 얻어지는 모든 잡음에 대한 변수에 의해 각각의 하부 모델들을 특성화시켰다. 표적의 기동중에 나타나는 가속도를 효과적으로 다루기 위하여, 잡음의 크기가 급격히 증가할 경우 증가분을 가속도로 인식하여 기동표적 관계식에 이용하였다. 또한 모르는 가속도에 따른 시변 변수를 적응적으로 어립잡기는 어렵기 때문에 정밀한 계산을 위하여 퍼지 뉴럴 네트워크와 적응 상호작용 다중모델 기법을 이용하였다. 퍼지 뉴럴 네트워크의 동정을 위해서는 오차 역전파 학습법을 사용하였다. 그리고 제안된 알고리즘의 수행 가능성을 보여주기 위하여 몇 가지 예를 제시하였다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2012.11a
• /
• pp.652-654
• /
• 2012

본 논문에서는 다중 심도 카메라 기반의 실시간 피플 카운팅 시스템을 제안 한다. 카메라 영상으로부터 사람을 감지하고 추적하는 시스템 및 그 방법에 관한 것으로, 피플 카운팅 시스템은 쇼핑몰이나 대형건물의 출입구 등과 같은 다양한 환경에 적용될 수 있다. 기존 피플 카운팅 시스템에서의 급격한 조명의 변화나 겹침 현상, 가림 현상에 대한 해결 방법으로, 다중 심도 카메라 환경에서 동일 객체 추적을 위해 RLM(Range Laser Method)를 적용하고, 조명 등 환경 변화에 강인한 배경 제거 및 물체 검출 기법으로 가우시안 혼합 모델(Gaussian Mixture Model)을 적용해 객체인식에 대한 정확도를 높인다. 또한, 객체를 블랍(Blob)으로 지정해 확장 칼만 필터(Extended Kalman Filter, EKF) 방법으로 객체를 추적한다. 본 제안은 피플 카운팅 시스템에의 객체 검출 및 인식에 대한 정확도를 향상시킬 수 있으리라 기대된다.

• Current Optics and Photonics
• /
• v.4 no.6
• /
• pp.530-539
• /
• 2020

Hyperspectral images feature a relatively narrow band and are easily disturbed by noise. Accurate estimation of the types and parameters of noise in hyperspectral images can provide prior knowledge for subsequent image processing. Existing hyperspectral-noise estimation methods often pay more attention to the use of spectral information while ignoring the spatial information of hyperspectral images. To evaluate the noise in hyperspectral images more accurately, we have proposed a spectral-spatial cooperative noise-evaluation method. First, the feature of spatial information was extracted by Gabor-filter and K-means algorithms. Then, texture edges were extracted by the Otsu threshold algorithm, and homogeneous image blocks were automatically separated. After that, signal and noise values for each pixel in homogeneous blocks were split with a multiple-linear-regression model. By experiments with both simulated and real hyperspectral images, the proposed method was demonstrated to be effective and accurate, and the composition of the hyperspectral image was verified.

• KIEAE Journal
• /
• v.8 no.4
• /
• pp.19-27
• /
• 2008

The carbon dioxide is brought from the energy consumption and regarded as a criteria material to estimate the Global Warming Potential. Building shares about 30% in national energy consumption and affects to environment as much as the energy consumption. But there is not enough data to forecast the amount of the carbon dioxide during the maintenance stage. Various factors are related with the energy consumption and carbon dioxide emission such as the physical area, the building exterior area, the maintenance type and location. Among these factors, the building carbon-dioxide emission can be estimated by the overall building characteristics such as the maintenance area, the number of household, the heating type, etc., The physical amount such as the thickness of the insulation and window infiltration could explained the limited scope and might not be use to estimate the total carbon-dioxide emission energy because the each value could not include or represent the overall building. In this paper, it provided the estimation model of the carbon-dioxide emission, explained by the overall building characteristics. These factors are shown as the maintenance area, no. of household, the heating type, the volume of the building, the ratio of the window to wall area etc., For providing the estimation model of th carbon-dioxide emission, it conducted the corelation analysis to filter the variables and suggested the estimation model with the power model and multiple regression model. Most of the model have a good statistics and fitted in the curve line.

• The Journal of Korea Robotics Society
• /
• v.4 no.1
• /
• pp.25-33
• /
• 2009

One of the main problems of topological localization in a real indoor environment is variations in the environment caused by dynamic objects and changes in illumination. Another problem arises from the sense of topological localization itself. Thus, a robot must be able to recognize observations at slightly different positions and angles within a certain topological location as identical in terms of topological localization. In this paper, a possible solution to these problems is addressed in the domain of global topological localization for mobile robots, in which environments are represented by their visual appearance. Our approach is formulated on the basis of a probabilistic model called the Bayes filter. Here, marginalization of dynamics in the environment, marginalization of viewpoint changes in a topological location, and fusion of multiple visual features are employed to measure observations reliably, and action-based view transition model and action-associated topological map are used to predict the next state. We performed experiments to demonstrate the validity of our proposed approach among several standard approaches in the field of topological localization. The results clearly demonstrated the value of our approach.

• ETRI Journal
• /
• v.46 no.2
• /
• pp.165-174
• /
• 2024

We propose a method to suppress the speckle noise and blur effects of the light field extracted from a hologram using a deep-learning technique. The light field can be extracted by bandpass filtering in the hologram's frequency domain. The extracted light field has reduced spatial resolution owing to the limited passband size of the bandpass filter and the blurring that occurs when the object is far from the hologram plane and also contains speckle noise caused by the random phase distribution of the three-dimensional object surface. These limitations degrade the reconstruction quality of the hologram resynthesized using the extracted light field. In the proposed method, a deep-learning model based on a generative adversarial network is designed to suppress speckle noise and blurring, resulting in improved quality of the light field extracted from the hologram. The model is trained using pairs of original two-dimensional images and their corresponding light-field data extracted from the complex field generated by the images. Validation of the proposed method is performed using light-field data extracted from holograms of objects with single and multiple depths and mesh-based computer-generated holograms.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.5
• /
• pp.461-467
• /
• 2020

In noisy and multi-speaker environments, the performance of speech recognition is unavoidably lower than in a clean environment. To improve speech recognition, in this paper, the signal of the speaker of interest is extracted from the mixed speech signals with multiple speakers. The VoiceFilter model is used to effectively separate overlapped speech signals. In this work, clustering by Probabilistic Linear Discriminant Analysis (PLDA) similarity score was employed to detect the speech signal of the interested speaker, which is used as the reference speaker to VoiceFilter-based separation. Therefore, by utilizing the speaker feature extracted from the detected speech by the proposed clustering method, this paper propose a speaker diarization system using only the mixed speech without an explicit reference speaker signal. We use phone-dataset consisting of two speakers to evaluate the performance of the speaker diarization system. Source to Distortion Ratio (SDR) of the operator (Rx) speech and customer speech (Tx) are 5.22 dB and -5.22 dB respectively before separation, and the results of the proposed separation system show 11.26 dB and 8.53 dB respectively.

• Journal of Environmental Science International
• /
• v.23 no.12
• /
• pp.2045-2056
• /
• 2014

In this study, pressure drop was measured in the pulse jet bag filter without venturi on which 16 numbers of filter bags (Ø$140{\times}850{\ell}$) are installed according to operation condition(filtration velocity, inlet dust concentration, pulse pressure, and pulse interval) using coke dust from steel mill. The obtained 180 pressure drop test data were used to predict pressure drop with multiple regression model so that pressure drop data can be used for effective operation condition and as basic data for economical design. The prediction results showed that when filtration velocity was increased by 1%, pressure drop was increased by 2.2% which indicated that filtration velocity among operation condition was attributed on the pressure drop the most. Pressure was dropped by 1.53% when pulse pressure was increased by 1% which also confirmed that pulse pressure was the major factor affecting on the pressure drop next to filtration velocity. Meanwhile, pressure drops were found increased by 0.3% and 0.37%, respectively when inlet dust concentration and pulse interval were increased by 1% implying that the effects of inlet dust concentration and pulse interval were less as compared with those changes of filtration velocity and pulse pressure. Therefore, the larger effect on the pressure drop the pulse jet bag filter was found in the order of filtration velocity($V_f$), pulse pressure($P_p$), inlet dust concentration($C_i$), pulse interval($P_i$). Also, the prediction result of filtration velocity, inlet dust concentration, pulse pressure, and pulse interval which showed the largest effect on the pressure drop indicated that stable operation can be executed with filtration velocity less than 1.5 m/min and inlet dust concentration less than $4g/m^3$. However, it was regarded that pulse pressure and pulse interval need to be adjusted when inlet dust concentration is higher than $4g/m^3$. When filtration velocity and pulse pressure were examined, operation was possible regardless of changes in pulse pressure if filtration velocity was at 1.5 m/min. If filtration velocity was increased to 2 m/min. operation would be possible only when pulse pressure was set at higher than $5.8kgf/cm^2$. Also, the prediction result of pressure drop with filtration velocity and pulse interval showed that operation with pulse interval less than 50 sec. should be carried out under filtration velocity at 1.5 m/min. While, pulse interval should be set at lower than 11 sec. if filtration velocity was set at 2 m/min. Under the conditions of filtration velocity lower than 1 m/min and high pulse pressure higher than $7kgf/cm^2$, though pressure drop would be less, in this case, economic feasibility would be low due to increased in installation and operation cost since scale of dust collection equipment becomes larger and life of filtration bag becomes shortened due to high pulse pressure.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.9
• /
• pp.139-147
• /
• 2014

In this paper, we present a novel method for simultaneous detection and tracking of multiple objects using dual-layer particle filtering. The proposed dual-layer particle sampling (DLPS) algorithm consists of parent-particles (PP) in the first layer for detecting multiple objects and child-particles (CP) in the second layer for tracking objects. In the first layer, PPs detect persons using a classifier trained by the intersection kernel support vector machine (IKSVM) at each particle under a randomly selected scale. If a certain PP detects a person, it generates CPs, and makes an object model in the detected object region for tracking the detected object. While PPs that have detected objects generate CPs for tracking, the rest of PPs still move for detecting objects. Experimental results show that the proposed method can automatically detect and track multiple objects, and efficiently reduce the processing time using the sampled particles based on motion distribution in video sequences.

• Smart Media Journal
• /
• v.5 no.3
• /
• pp.35-41
• /
• 2016

In this paper, we propose hand-gesture signal recognition based on EPS(Electronic Potential Sensor) using Deep learning model. Extracted signals which from Electronic field based sensor, EPS have much of the noise, so it must remove in pre-processing. After the noise are removed with filter using frequency feature, the signals are reconstructed with dimensional transformation to overcome limit which have just one-dimension feature with voltage value for using convolution operation. Then, the reconstructed signal data is finally classified and recognized using multiple learning layers model based on deep learning. Since the statistical model based on probability is sensitive to initial parameters, the result can change after training in modeling phase. Deep learning model can overcome this problem because of several layers in training phase. In experiment, we used two different deep learning structures, Convolutional neural networks and Recurrent Neural Network and compared with statistical model algorithm with four kinds of gestures. The recognition result of method using convolutional neural network is better than other algorithms in EPS gesture signal recognition.