• 한국정보통신학회논문지
• /
• 제21권3호
• /
• pp.629-635
• /
• 2017

본 논문에서는 영상에 첨가된 가우시안 잡음과 임펄스 잡음의 영향을 완화하기 위하여 잡음의 종류에 따라 처리하는 비선형 합성 필터를 제안하였다. 잡음 판단을 통해 국부 마스크의 중심화가 가우시안 잡음으로 판단된 경우, 국부 마스크 내의 표본분산을 이용하여 공간 가중치 필터와 화소 변화에 따른 가중치 필터의 가중치를 다르게 적용하여 처리하고, 임펄스 잡음으로 판단된 경우, 국부 마스크의 잡음 밀도에 따라 국부 히스토그램 가중치 필터와 표준 메디안 필터의 가중치를 다르게 적용하여 처리하는 알고리즘을 제안하였다. 그리고 제안한 필터 알고리즘의 성능을 평가하기 위해 PSNR(peak signal to noise ratio)을 사용하여 기존의 방법들과 제안한 필터 알고리즘을 가우시안 잡음, 임펄스 잡음 및 두 잡음이 혼합된 복합잡음 환경에서 각각 비교하였다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제16권11호
• /
• pp.3507-3522
• /
• 2022

Model pruning methods have attracted huge attention owing to the increasing demand of deploying models on low-resource devices recently. Most existing methods use the weight norm of filters to represent their importance, and discard the ones with small value directly to achieve the pruning target, which ignores the contribution of the small norm filters. This is not only results in filter contribution waste, but also gives comparable performance to training with the random initialized weights [1]. In this paper, we point out that the small norm filters can harm the performance of the pruned model greatly, if they are discarded directly. Therefore, we propose a novel filter contribution recycle (FCR) method for structured model pruning to resolve the fore-mentioned problem. FCR collects and reassembles contribution from the small norm filters to obtain a mixed contribution collector, and then assigns the reassembled contribution to other filters with higher probability to be preserved. To achieve the target FLOPs, FCR also adopts a weight decay strategy for the small norm filters. To explore the effectiveness of our approach, extensive experiments are conducted on ImageNet2012 and CIFAR-10 datasets, and superior results are reported when comparing with other methods under the same or even more FLOPs reduction. In addition, our method is flexible to be combined with other different pruning criterions.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2013년도 춘계학술대회
• /
• pp.911-913
• /
• 2013

사용자의 행동 추론에서는 수신된 센서 데이터의 신뢰성이 중요하다. 하지만 외 내부 환경, 온도, 진동 등에 따라 센서의 측정값이 달라진다. 이러한 잡음 환경은 센서로부터 수집되는 정보의 신뢰성에 영향을 준다. 때문에 잡음을 감소시키고 사용자의 행동을 추론하는 디지털 필터와 추론 알고리즘이 요구된다. 본 연구는 사용자 행동의 잡음 제거, 왜곡 방지로 칼만 필터와 지수 가중이동 평균 필터의 비교실험을 하였다. 또한 본 제안된 확장가중 이중필터 법과 비교하였다. 결과적으로 제안된 필터가 다른 필터에 비해 잡음 환경에 견고하였다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권10호
• /
• pp.3534-3553
• /
• 2021

The low cost of Bluetooth technology has led to its wide usage in indoor positioning. However, some inherent shortcomings of Bluetooth technology have limited its further development in indoor positioning, such as the unstable positioning state caused by the fluctuation of Received Signal Strength Indicator (RSSI) and the low transmission frequency accompanied by a poor real-time performance in positioning and tracking moving targets. To address these problems, an indoor fusion positioning algorithm of Bluetooth technology and pedestrian dead reckoning (PDR) based on a particle filter with dynamic adjustment of weights calculation strategy (BPDW) will be proposed. First, an orderly statistical filter (OSF) sorts the RSSI values of a period and then eliminates outliers to obtain relatively stable RSSI values. Next, the Group-based Trilateration algorithm (GTP) enhances positioning accuracy. Finally, the particle filter algorithm with dynamic adjustment of weight calculation strategy fuses the results of Bluetooth positing and PDR to improve the performance of positioning moving targets. To evaluate the performance of BPDW, we compared BPDW with other representative indoor positioning algorithms, including fingerprint positioning, trilateral positioning (TP), multilateral positioning (MP), Kalman filter, and strong tracking filter. The results showed that BPDW has the best positioning performance on static and moving targets in simulation and actual scenes.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2000년도 제15차 학술회의논문집
• /
• pp.357-357
• /
• 2000

In this study, we consider a multi-input multi-output styrene polymerization reactor system for which the monomer conversion and the weight average molecular weight are controlled by manipulating the jacket inlet temperature and the feed flow rate. The reactor system is identified by using a linear subspace identification method and then the output feedback model predictive controller is constructed on the basis of the identified model. Here we use the Best Linear Unbiased Estimation (BLUE) filter as a stochastic estimator instead of the Kalman filter. The BLUE filter observes the state successfully without any a priori information of initial states. In contrast to the Kalman filter, the BLUE filter eliminates the offset by observing the state of the augmented system regardless of a priori information of the initial state for an integral white noise augmented system. A BLUE filter has a finite impulse response (FIR) structure which utilizes finite measurements and inputs on the most recent time interval [i-N, i] in order to avoid long processing times.

• 한국정밀공학회지
• /
• 제31권10호
• /
• pp.905-911
• /
• 2014

A Loss-in-Weight (LIW) feeder, a type of automated measuring device, is a continuous feeder used in many mass production industries. Due to its versatility, there have been constant demands of LIW feeders in food production supply lines as well as chemical and pharmaceutical industries. In this paper, the process of designing a LIW feeder system with better performance will be examined and compared with commercial products. This system is characterized by low pass Butterworth filter and feed forward PI control. The filter is for noise disposal caused by dynamic condition of a LIW feeder. The feed forward PI control, based on linearity feature of feeders, is adequate for stable driving of the system. At the end, a possible evaluation method of LIW system will be proposed to verify the specific achievement of this paper.

• 대한전기학회논문지:시스템및제어부문D
• /
• 제50권6호
• /
• pp.302-308
• /
• 2001

SVD(Singular Value Decomposition) filter for the suppression of EMG in multi-lead stress ECG is studied. SVD filter consists of two parts. In the first part, the basis vectors were chosen from the averaged singular vectors obtained from the decomposed noise-free ECG. The singular vector is computed from the stress ECG and is compared itself with basis vectors to know whether the noise exist in stress ECG. In the second part, the existing elimination method is used, when one(or two) channels is(or are) contaminated by noise. But the proposed enhanced SVD filter is used in case of having the noise in the many channels. During signal decomposition and reconstruction, the noise-free channel or the least noisy channel have the weight of 1, the next less noisy channel has the weight of 0.8. In this way, every channel was weighted by decreased of 0.2 in proportion to the amount of the added noise. For the evaluation of the proposed enhanced SVD filter, we compared the SNR computed by the enhanced SVD filter with the standard average filter for the noise-free signal added with artificial noise and the patient data. The proposed SVD filter showed better in the SNR than the standard average filter. In conclusion, we could find that the enhanced SVD filter is more proper in processing multi-lead stress ECG.

• 전자공학회논문지S
• /
• 제36S권7호
• /
• pp.81-89
• /
• 1999

통신에 의한 전송 영상은 잡음이나 번짐 또는 일그러짐 등을 항상 포함한다. 본 논문에서는 적응형 일반스텍 최적화 필터(OAGSF: optimal adaptive generalized stack filter)라는 영상복원 공간 필터를 제안하였는데, 이는 영상의 복원에서 잡음 제거율과 외곽선 정보의 보존률의 증가을 위해 신경회로맘의 역전파 학습 알고리즘의 가중치 학습 알고리즘을 기반으로 적응형 일반스택 필터(AGSF)를 최적화 시킨 것이다. 적응형 일반스택 필터는 일반스택 필터(GSF: generalized stack filter)와 적응형 다단계 메디안 필터(AMMF; adaptive multistage median filter)로 구분하고, 일반스텍 필터는 스택 필너치 기능을 보완한것이고, 적응형 다단계 메디안 필터는 메디안 필터의 외곽선 정보 보존률을 높인 것이다. 신경회로망의 역전파 학습 알고리즘에 대하여 두가지 가중치 학습 알고리즘인 최소평균절대 (LMA:Least Mean Absolute) 알고리즘과 최소평균자승(LMS: Least Mean Square) 알고리즘을 이용하여 적응형 일반스택 필터를 최적화하였다. 본 논문에서 제시한 신경회로망을 이용한 영상복원 공간필터에 대해 실험결과를 통해 제시하였다.

• 한국연초학회지
• /
• 제23권2호
• /
• pp.179-184
• /
• 2001

This study was conducted to determine the smoke component distribution and the relationship between chemical components and physical characteristics of cigarettes. 16 different cigarette brands which were sold in the market were selected for this study. Five kinds of smoke components which have been tar, nicotine, water, carbon monoxide(CO) puff No., and six kinds of physical characteristics which were filter type, leaf weight, filter weight, UPD, EPD, dilution rate were analyzed. The average values in tar, nicotine, water, CO concentration were 6.5 mg/cig. 0.66 mg/cig, 1.12 mg/cig. and 6.32 mg/cig., respectively. The average ratios of nicotine/tar and CO/tar were 0.10, and 1.02 respectively. The distribution of smoke components collected in the cambridge filter and cigarette filter was different. The averages of tar and nicotine removal efficiency by a cigarette filter were 53%, and 48%, respectively. All smoking components were positively correlated with other smoking components. filter types, EPD, and dilution rate were showed high correlation to the changes of smoke components. Especially, dilution rate of cigarette strongly affected on the changes of all smoke components.

• Plant Resources
• /
• 제7권1호
• /
• pp.77-86
• /
• 2004

The reduced ratio of red:far-red (R:FR) light acts as a measure of the proximity of competitors and plants can detect the potentially competing neighbor plants by perceiving reflected R:FR signals and initiate the response of “shade avoidance” before actual shading occurs. The phytochrome system is responsible for monitoring the changes in the R:FR and initiating the shade avoidance response. The response to low R:FR ratio was studied in a white aspen Populus alba clone ‘Bolleana’ using two filter systems: a clear plastic filter system that allows a R:FR ratio less than 1.0 to pass from adjacent border plant reflection; and a special commercial plastic that blocks FR light and creates a R:FR ratio above 3.0. The reduced R:FR signals enhanced the stem elongation in response to competition at the expense of relative stem diameter growth. Trees grown inside clear chambers were 27 ％ taller than trees grown inside the FR-blocking filter chambers. Stem taper of clear chamber trees was 16％ less than the FR-blocking filter trees. Low R:FR also induced 22％ more stem dry weight and 13％ greater petiole length per leaf compared to the FR-blocking filter trees. There were no statistically significant differences in leaf area, leaf number increment, and total dry weight between the two light filter treatments.