Abstract:
PURPOSE: A data selection method in an object tracking system based on gradual learning is provided to effectively learn the outer appearance model of an object. CONSTITUTION: A kind information assigning unit(321) presently assigns kind information to tracking data and comparison tracking data of the update stage. A probability calculation unit(322) calculates the probability of the tracking data and the kind information tracking data located in the local domain of the comparison tracking data. An entropy calculating unit(323) calculates the entropy in the local domain of the tracking data and the comparison tracking data. A data sorter(324) classifies the tracking data into kind boundary data and kind non-boundary data by using the calculated entropy.
Abstract:
A fault detection method using sequential one class classifier chain is provided to detect an abnormality which is generated in a process by using a sensor data monitoring a process. A fault detection method using sequential one class classifier chain is comprised of the steps: a normal process of facility is monitored by a learning preprocess phase of sensor data(ST110); a learning feature extraction step reduces the dimension of the pre-processed sensor data as described above(ST120); The critical data group selection stage is provided selection toward data of the importance group(ST130); and after leaning is performed by a selected data group and a classification unit is configured in a detailed classification constitutional state(ST140).
Abstract:
본 발명은 환자의 다양한 임상 정보를 입력하여 전립선암으로 의심되는 확률을 출력함으로써 조직 검사 시행 여부 및 전립선암 발병 진단에 도움을 줄 수 있는 암 진단 방법 및 장치를 제공하는 것을 목적으로 한다. 이를 위하여, 본 발명은 하나 이상의 대상자의 임상 정보 수치 및 암 진행 단계 수치를 포함하는 학습 데이터를 분류기에 입력하는 단계; 분류기에서 상기 학습 데이터를 이용하여 의사결정 하이퍼플레인(decision hyperplane)을 생성하는 단계; 검사 대상자의 임상 정보 수치를 분류기에 입력하는 단계; 분류기에서 입력된 검사 대상자의 임상 정보 수치를 이용하여 검사 대상자의 암 진행 단계 수치를 출력하는 단계; 분류기에서 상기 출력된 암 진행 단계 수치를 의사결정 하이퍼플레인을 기준으로 분류하는 단계; 및 상기 분류기의 출력된 암 진행 단계 수치를 이용하여 검사 대상자가 암에 걸렸을 확률을 계산하는 단계를 포함하는 암 진단 방법 및 그에 사용가능한 진단 장치를 제공한다.
Abstract:
PURPOSE: A method and apparatus for diagnosing prostate cancer are provided to reduce unnecessary tissue test and side effects. CONSTITUTION: A method for diagnosing prostate cancer comprises: a step of putting learning data including clinical information value and cancer progression value in a sorter; a step of generating decision hyperplane in the sorter using learning data; a step of inputting clinical information value in the sorter; a step of outputting cancer progression value of tester using the clinical information value of the tester; a step of classifying output cancer progress based on decision hyperplane; and a step of calculating probability of prostate cancer of tester.
Abstract:
본 발명에 따른 제어 대상물의 출력값과 목표값의 차이에 근거하여 제어 대상물에 제어값을 제공하는 비례-적분-미분(proportional-integrate-derivative: PID) 제어 장치는, 목표값에 대해 상기 제어 대상물의 출력값이 갖는 오차값을 출력하는 오차 산출기와; 상기 오차값에 대해 비례 연산을 수행하여 비례 연산값을 얻고, 상기 오차값에 대해 적분 연산을 수행하여 적분 연산값을 얻으며, 상기 오차값에 대해 미분 연산을 수행하여 미분 연산값을 얻고, 상기 비례 연산값, 적분 연산값 및 미분 연산값에 근거하여 얻어진 제어값을 상기 대상물로 출력하는 PID 연산기와; 하나의 목표값에 대해 상기 제어 대상물의 출력값을 복수회 샘플링하여 상기 오차 산출기로 출력하는 제1 샘플러와; 상기 제1 샘플러의 샘플링 주기에 따라서 상기 PID 연산기가 복수회 PID 연산을 수행하여 제어값을 출력하도록 제어하는 제어부를 포함한다. PID 제어, 퍼지 제어, 비례 계수, 샘플러, 멀티레이트
Abstract:
A PID control apparatus and a method thereof are provided to minimize the configuration of circuit by utilizing a multi-rate technique. A PID(Proportional-Integrate-Derivative) control apparatus includes an error calculator(220), a PID calculator(230), a first sampler(240), and a controller(260). The error calculator outputs an error value of an output of a controlled object on a target value. The PID calculator obtains a proportional calculation value by performing a proportional calculation on the target value, obtains an integral calculation value by performing an integral calculation, obtains a differential calculation value by performing a differential calculation, and outputs a control value, which is obtained based on the proportional, integral, and differential calculations, to the controlled object. The first sampler samples the output of the controlled object on one target value and outputs the sampled value to the error calculator. The controller controls the PID calculator in order to output a control value by performing a PID calculation many times according to a sampling period of the first sampler.
Abstract:
An apparatus for correcting hand-shaking of a camera module for mobile appliances is provided to form an image at the same position of an image sensor by controlling the position of the image sensor according to the amount and direction of hand-shaking when taking a picture. An apparatus for correcting hand-shaking of a camera module for mobile appliances comprises a camera unit(100) and an OIS(Optical Image Stabilizer) circuit unit(200). The camera unit includes an angular speed sensor(110) for sensing the angular speed of hand-shaking of a camera, a position detecting sensor(120) for sensing the present position of an image sensor and a drive unit(130) driving the image sensor. The OIS circuit unit controls the control unit with a multi-rate PID control method using a relatively little control period compared to conventional PID control.
Abstract:
본 발명은 점진적 학습 기반의 물체 추적 시스템에서의 데이터 선택 방법 및 장치에 관한 것으로, 본 발명의 일 실시 예에 따른 방법은 외양 모델을 갱신하기 위한 추적 데이터들과 비교 추적 데이터들을 비교하고, 비교 결과에 따라 상기 추적 데이터들 중에서 일부를 선택함으로써, 외양 모델 갱신에 보다 적합한 추적 데이터를 선택하여 추적 성능을 향상시킨다. 점진적 학습, 물체 추적, 외양 모델, 선택
Abstract:
PURPOSE: A concept learning method using an evolutionary logic network with a learning result is provided to enhance a concept learning speed by reusing a learned concept evolutionary accumulated in the concept logic network and using the learning result previously acquired for other concept. CONSTITUTION: An optimal property node for concept learning data is selected from the logic network. A basic operation executed by corresponding to a property node selection result is selected. In case that the optimal property node is selected, a node level operation is selected as the basic operation to be executed. If not, a network level operation is selected as the basic operation to be executed.
Abstract:
PURPOSE: Methods for obtaining a model, an estimated value and a compensated value of a thermal transformation are provided to improve the processing accuracy by compensating the thermal transformation. CONSTITUTION: A temperature vector is obtained by measuring temperature information from a processing device(110). A backmix vector function to set a relation between an estimated value vector of a heating source and the temperature vector is calculated. The estimated value vector of the heating source is obtained by using the backmix vector function. Important estimated value variables of the heating source are divided from the estimated value vector of the heating source. Important temperature variables are divided from the temperature vector. A model formula of a thermal transformation to show the relation between the important temperature variables and an estimated value of a thermal transformation is set. Thereby, the optimized model of the thermal transformation is obtained. Moreover, the accuracy of the processing device is improved by effectively removing errors for the thermal transformation.