• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.12 no.6
• /
• pp.629-635
• /
• 2019

In Deep Learning models derivative is implemented by error back-propagation which enables the model to learn the error and update parameters. It can find the global (or local) optimal points of parameters even in the complex models taking advantage of a huge improvement in computing power. However, deliberately generated data points can 'fool' models and degrade the performance such as prediction accuracy. Not only these adversarial examples reduce the performance but also these examples are not easily detectable with human's eyes. In this work, we propose the method to detect adversarial datasets with random noise addition. We exploit the fact that when random noise is added, prediction accuracy of non-adversarial dataset remains almost unchanged, but that of adversarial dataset changes. We set attack methods (FGSM, Saliency Map) and noise level (0-19 with max pixel value 255) as independent variables and difference of prediction accuracy when noise was added as dependent variable in a simulation experiment. We have succeeded in extracting the threshold that separates non-adversarial and adversarial dataset. We detected the adversarial dataset using this threshold.

• IEIE Transactions on Smart Processing and Computing
• /
• v.3 no.5
• /
• pp.285-297
• /
• 2014

This paper presents the results of an investigation into bi-directional communication in spectrum sharing-based cognitive radio (Bi-CR) systems. A Bi-CR system can increase the spectral efficiency significantly by sharing the spectrum and through the bi-directional use of spatial resources for two-way communication. On the other hand, the primary user experiences more interference from the secondary users in a Bi-CR system. Satisfying the interference constraint by simply reducing the transmission power results in performance degradation for secondary users. In addition, secondary users also experience self-interference from echo channels due to full duplexing. These imperfections may weaken the potential benefits of the Bi-CR system. Therefore, a new way to overcome these defects in the Bi-CR system is needed. To address this need, this paper proposes some novel power allocation schemes for the Bi-CR system. This contribution is based on two major analytic environments, i.e., noise-limited and interference-limited environments, for providing useful analysis. This paper first proposes an optimal power allocation (OPA) scheme in a noise-limited environment and then analyzes the achievable sum rates. This OPA scheme has an effect in the noise-limited environment. In addition, a power allocation scheme for the Bi-CR system in an interference-limited environment was also investigated. The numerical results showed that the proposed schemes can achieve the full duplexing gain available from the bi-directional use of spatial resources.

• International Journal of Highway Engineering
• /
• v.17 no.2
• /
• pp.135-141
• /
• 2015

PURPOSES : While driving, drivers are usually limited in communicating with others except for using the horn. Excessive use of the horn may cause noise pollution, quarrels between drivers and pedestrians, damage, etc. This study developed a methodology for environmental adaptation and improvements of vehicle horns. METHODS: In this study, we performed a literature review of previous studies and related technologies regarding the overuse and damage of the horn. The proposed methodology employed the paired comparison method, as well as the semantic differential method. These methods can consider various vehicle horns, such as the Sport Utility Vehicle(SUV) Horn, Van Horn, and Buzzer. In addition, we conducted a factor analysis in order to provide a direction for improvement of future horns. This research provides a means of complimenting existing intellectual property related to vehicle horns. RESULTS: As a result, the most preferred of the selected horns were the Buzzer at 86.7%. In addition, based on the factor analysis, the horns could be classified into pleasantness and comfort factors. The results indicate a positive reaction for various vehicle horns. The horn type and size of the process control have been properly developed considering the position of the vehicle and the surrounding noise measurements. CONCLUSIONS : Based on the proposed methodology, public and private sectors can use fundamental data for reasonable traffic-noise control policies.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2007.11a
• /
• pp.1177-1182
• /
• 2007

This study aims at identifying the effect of preferred visual and auditory stimuli on the road traffic noise perception through the series of psychoacoustic experiments. The first experiment was designed to find the addition effect of a sound reproducing system in a square where lay adjacent to roads with road traffic noise, provide appropriate music corresponding to the varying conditions of weather in the given space. In result, it was found that the rating around the square improved toward more positive adjectives. The second experiment was designed to assess the visual effect of twelve roadsides with different landscapes on the road traffic noise perception. As a result, approx. 3 to 5 dB(A) of psychological reduction was seen in places where natural landscape was preferred, compared to the others, although the noise levels were similar. The third experiment was designed to evaluate the effect of visual screen from adjacent roads on road traffic noise perception by means of ME method in a laboratory. In result, the effect of psychological reduction was observed at 65dB(A) or lower. Especially, complete screening from adjacent roads led to 5 to 10% of loudness reduction effect, compared to non-screening cases. Finally, the fourth experiment was designed to evaluate the effect of visual and auditory information with ME method and 7-point SD rating scale in a laboratory. In result, up to 10% of loudness reduction and about 2dB(A) of noise perceptional reduction were seen at 65dB(A) or lower.

• The Journal of the Acoustical Society of Korea
• /
• v.26 no.5
• /
• pp.220-226
• /
• 2007

The two major objectives of acoustical measurements in a cavitation tunnel are measuring the noise levels generated by rotating propellers behind a hull and localizing possible noise sources in order to reduce noise levels. Propeller noise measurement experiments were performed in MOERI cavitation tunnel at December, 2006. In order to put the propeller into cavitating conditions, a wake-generating dummy body was devised. In addition, ten hydrophones are put inside a wing-shaped casing in order to minimize the unexpected flow induced self noise around hydrophone itself. After measuring both of the noises of the rotating propeller behind the dummy body and signals generated by a virtual source, respectively the data were matched field processed using the frequency incoherent Bartlett processor to localize noises on the propeller plane. In this paper, we presented the measured noise analysis and the localization results.

• Proceedings of the IEEK Conference
• /
• 2000.06e
• /
• pp.183-186
• /
• 2000

In this paper we propose a new architecture of a programmable digital automatic gain controller(AGC) for analog interface in mixed mode systems. Compared with conventional analog AGCs which have difficulties in integration due to large capacitors, the proposed AGC is easily integrated. So the production cost can be reduced. In addition, The proposed AGC has a better performance in temperature, and power supply variations, and substrate noise than analog counterparts do. To prevent erroneous operations of the AGC due to noise, a mal-function preventer is newly proposed. In addition, to achieve an optimized AGC time constant, we propose a logic block which controls an up-down counting clock. This is directly related to the changing speed of the AGC gain. Implemented with a 0.25 $\mu\textrm{m}$ 1-poly, 5-metal CMOS parameters, the AGC operates from a single 2.5V power supply with the dynamic range of 36.ldB and occupies active area of 500$\mu\textrm{m}$${\times}$600$\mu\textrm{m}$

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2004.11a
• /
• pp.283-286
• /
• 2004

This work presents a 3-axis active mount featuring the piezoelectric stack actuators. The inertia configuration for the piezoelectric actuators is adopted to achieve appropriate force and moment in each axis. In addition, the rubber mount is integrated to provide stiffness for system loading. In order to analyze the active mount, the governing equation for the force and moment is derived and its validity is demonstrated by comparing generated force and moment of each actuator between analysis and experiment. In addition, the natural frequency of each actuator is identified though both simulation and experiment.

• Proceedings of the IEEK Conference
• /
• 2000.11a
• /
• pp.197-200
• /
• 2000

In this paper, The RF system of W-CDMA receiver is designed and the performance is analyzed. The linearity characteristic and the noise characteristic are presented in the performance. The linearity characteristic is analyzed by PN and IIP3. The noise characteristic is analyzed by NF. In addition, sweeping of the nonlinear components parameter affecting the linear performance is tested and the most maximal possible parameter to maintain the linear performance is introduced. The transceiver RF system of W-CDMA and cdma2000 is designed and presented adapting the nonlinear parameter introduced.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2002.05a
• /
• pp.818-823
• /
• 2002

Addition of point supports results in increasing the fundamental frequency of a structure, generally. In this paper, searching more effective location of point supports is a major object to maximize a fundamental frequency of various cantilever plates. Results are presented by aspect ratio of the plate, by design domain within which point supports generate, and by mass location equipped on the plate. Optimization method is applied due to expand the ESO(Evolutionary Structural Optimization) method.

• Journal of KSNVE
• /
• v.2 no.2
• /
• pp.117-123
• /
• 1992

In the study, the design method of reconstruction filter is studied for synchronized sampling which is necessary for order tracking in rotating machinery. The original data sampled at constant intervals, using fixed anti- aliasing filters, is reconstructed by digital reconstruction filter and is resampled at new sampling times calculated by a suitable shaft angle encoder pulse arrival times in order to synchronize with shaft velocity. In addition to eliminating the tracking synthesizer and filters, this new method has no phase noise due to phase-locked loops.