• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.7 no.12
• /
• pp.3274-3292
• /
• 2013

Unsupervised methods for image segmentation are recently drawing attention because most images do not have labels or tags. A topic model is such an unsupervised probabilistic method that captures latent aspects of data, where each latent aspect, or a topic, is associated with one homogeneous region. The results of topic models, however, usually have noises, which decreases the overall segmentation performance. In this paper, to improve the performance of image segmentation using topic models, we propose two topic masks applicable to topic assignments of homogeneous regions obtained from topic models. The topic masks capture the noises among the assigned topic assignments or topic labels, and remove the noises by replacements, just like image masks for pixels. However, as the nature of topic assignments is different from image pixels, the topic masks have properties that are different from the existing image masks for pixels. There are two contributions of this paper. First, the topic masks can be used to reduce the noises of topic assignments obtained from topic models for image segmentation tasks. Second, we test the effectiveness of the topic masks by applying them to segmented images obtained from the Latent Dirichlet Allocation model and the Spatial Latent Dirichlet Allocation model upon the MSRC image dataset. The empirical results show that one of the masks successfully reduces the topic noises.

• Transactions of the Korean Society for Noise and Vibration Engineering
• /
• v.24 no.10
• /
• pp.802-808
• /
• 2014

The purpose of this study is to suggest efficient methods for estimating population exposed to noises by analyzing differences of population exposed to noises estimated by each method through comparing exposed population estimated by utilizing existing methods and those estimated by using census output areas reflecting the actual population information of each address. For population exposed to noises, the error of exposed population estimated by using the per capita living space turned out to be the biggest, and other estimation methods had no significant difference. For population exposed to excess noises, as a result of analyzing population estimated by each method based on census output areas, the error of the method using a grid noise map turned out to be the biggest. For the method to estimate population exposed to noises by using a noise map, the estimation methods using census output areas and total ground area are considered to be more rational than the grid noise map estimation method or the method to estimate the living space per capita.

• Proceedings of the KSME Conference
• /
• 2001.06b
• /
• pp.187-191
• /
• 2001

Sound quality engineering in automobile noise applications has become more and more important under the current quiet driving condition that the interior noise level is below 65 dBA, because various noise components masked under high noise level can be audible in quieter driving situation. Many researches have been carried out for subjective and objective assessments on automobile sounds and noises. In particular, the interior sound quality has been one of research fields that can give high-quality feature to automobile products. Although many works related to the interior sound quality have been progressed or completed in foreign countries, limited research results are presented in the country. In the study, as a base step necessary to objective assessments on car interior noises, subjective assessments are performed with 20 subjects. For this purpose, perceptual adjectives suitable to the assessment of acceleration noises are selected as assessment scales through questionnaire procedures using 35 subjects. Mean values and standard deviations are calculated for noises created through digital filtering of acceleration noises measured. In addition, the correlation analysis and the factor analysis are carried out to investigate the dependence of the assessment scales selected.

• Journal of the Institute of Convergence Signal Processing
• /
• v.5 no.2
• /
• pp.106-110
• /
• 2004

The noise characteristics of the submarines and vessels are composed of tonal signals and background noises. Tonal signals are from the revolution of the motors or engines, and background noises are from the generation and disappearing of the bubbles around the vessels. We analyze the noises from maneuvering underwater vessels with ATW algorithm. We design broadband asymptotic noise generator using Hilbert transform and design the system which can generate asymptotic noise with sum of generated background noise and synthesized tonal signals. This system can be used as a cheating system from the torpedo shacks.

• Proceedings of the KIEE Conference
• /
• 2003.07c
• /
• pp.1845-1848
• /
• 2003

PD(Partial Discharge) signal have the range from several kHz to several GHz. External noises don't give any interference to discerning PD signals with high frequency bandwidth between several hundreds MHz to several GHz but, in the range between several decades kHz to several MHz, are big obstacles to prevent from discriminating PD signals. Therefore, selecting appropriate filter to decrease the influence of noise is important problem. PD signals have the characteristics both short duration and non-continuous pulse train. For separating PD signals from External noises, some different filtering methods different from conventional ones are needed and wavelet filtering method was convinced to be suitable filtering method in various investigations. This paper has simulated PD signal and various noises and investigated the influences for the performance of wavelet filter by varying the components relative to designing of wavelet filter and constructed suitable one. For the purpose of convincing the performance of wavelet filter, real PDs were produced in needle to plane electrode structure. This small PD signal was mixed with external noises and filtered with wavelet filter. We have obtained apparent PD signals and noises are well suppressed by wavelet filter.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2005.11a
• /
• pp.363-366
• /
• 2005

Social survey and auditory experiment on residential noises such as floor impact, air-bome, bathroom, drainage and traffic noises were conducted to investigate the effects of sensitivity and background noise levels on subjective response to residential noises. Results of survey showed that multiple questions for sensitivity to noise have a more power for discrimination against subjective response to noise than single question. It was also found that the correlation between sensitivity and annoyance was relatively low, but the differences in annoyance between sensitivity groups were significant. Result of auditory experiment showed that effects of sensitivity on annoyance were different according to noise type and noise level. It was also shown that the difference in annoyance between the lowest sensitive stoup and the highest sensitive group was significant in low noise level. The effects of background noise on subjective responses to noise was also different according to noise source and noise level, and were significant in low noise level of airborne and drainage noise.

• 제어로봇시스템학회:학술대회논문집
• /
• 1997.10a
• /
• pp.1356-1359
• /
• 1997

There are various methods to detect the faulty electric facilities(esp, insulator) indirectly on power distribution lines at a certain distance apart. This paper describes the proto type equipment to detect a faulty insulator by receiving a electric discharge noises, which are generated with a periodicity of 120Hz.

• 제어로봇시스템학회:학술대회논문집
• /
• 1993.10a
• /
• pp.456-459
• /
• 1993

CADET is used to analyze the performance of the missile. Miss distance is calculated for a given lateral fin force saturation level due to the aerodynamic characteristics, target acceleration, and glint and fading noises which is assumed as Gaussian noises. As .alpha.-.betha. filter is studied to attenuate the noises, the results are compared with those of without filter. For the easy simulation, the transfer function of a discrete .alpha.-.betha. filter is converted into the continuous model. Simulation results show that the results of CADET simulation is similar to those of Monte-Carlo simulation. Moreover CADET is the better in computing time demand.

• Proceedings of the Korean Institute of IIIuminating and Electrical Installation Engineers Conference
• /
• 2000.11a
• /
• pp.231-234
• /
• 2000

The corona in oil, the surface discharge in oil, the void discharge which could be in a transformer were imitated by using three electrodes with different shapes and the frequency bands of each ultrasonic signal were measured. After emitting artificial noises toward the transformer, only ultrasonic signals were extracted by frequency filtering, from the ultrasonic signals with noises at partial discharge. As a result, it was confirmed that the ultrasonic signals due to the corona in oil, the surface discharge in oil, the void discharge could be distinguished from the noises.

• The Journal of the Acoustical Society of Korea
• /
• v.28 no.1E
• /
• pp.16-22
• /
• 2009

In this paper, the new robust active noise controller was proposed to be applied for attenuating the noises when the nonlinear distortion in the secondary path exists. Through computer simulations as well as the analytical analysis, it could be shown that it is possible for both conventional LMS controller and proposed controller, to be applied for actively controlling the noises and linearizing the nonlinear distortion in the secondary path. Also, the simulations results demonstrated that the proposed controller may have faster convergence speed and better capability of controlling the noises and compensating the nonlinear distortion than the conventional LMS controller.