• KIPS Transactions on Software and Data Engineering
• /
• v.8 no.5
• /
• pp.193-204
• /
• 2019

The explosion of data due to the improvement of sensor technology and computing performance has become the basis for analyzing the situation in the industrial fields, and various attempts to detect events based on such data are increasing recently. In particular, sound signals collected from sensors are used as important information to classify events in various application fields as an advantage of efficiently collecting field information at a relatively low cost. However, the performance of sound-event classification in the field cannot be guaranteed if noise can not be removed. That is, in order to implement a system that can be practically applied, robust performance should be guaranteed even in various noise conditions. In this study, we propose a system that can classify the sound event after generating the enhanced sound signal based on the deep learning algorithm. Especially, to remove noise from the sound signal itself, the enhanced sound data against the noise is generated using SEGAN applied to the GAN with a VAE technique. Then, an end-to-end based sound-event classification system is designed to classify the sound events using the enhanced sound signal as input data of CNN structure without a data conversion process. The performance of the proposed method was verified experimentally using sound data obtained from the industrial field, and the f1 score of 99.29% (railway industry) and 97.80% (livestock industry) was confirmed.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.47 no.6
• /
• pp.162-170
• /
• 2010

Adaptive arrays reject interferences while preserving the desired signal, exploiting a priori information on its arrival angle. Subspace-based adaptive arrays, which adjust their weight vectors in the signal subspace, have the advantages of fast convergence and robustness to steering vector errors, as compared with the ones in the full dimensional space. However, the complexity of theses subspace-based methods is high because the eigendecomposition of the covariance matrix is required. In this paper, we present a simple subspace-based method based on the PASTd (projection approximation subspace tracking with deflation). The orignal PASTd algorithm is modified such that eigenvectora are orthogonal to each other. The proposed method allows us to significantly reduce the computational complexity, substantially having the same performance as the beamformer with the direct eigendecomposition. In addition to the simple beamforming method, we present theoretical analyses on the SINR (signal-to-interference plus noise ratio) of subspace beamformers to see their behaviors.

• The Journal of the Acoustical Society of Korea
• /
• v.21 no.6
• /
• pp.522-530
• /
• 2002

This paper introduces a neural network (NN) -based nonlinear predictor for the LP (Linear Prediction) residual. To evaluate the effectiveness of the NN-based nonlinear predictor for LP-residual, we first compared the average prediction gain of the linear long-term predictor with that of the NN-based nonlinear long-term predictor. Then, the effects on the quantization noise of the nonlinear prediction residuals were investigated for the NN-based nonlinear predictor A new NN predictor takes into consideration not only prediction error but also quantization effects. To increase robustness against the quantization noise of the nonlinear prediction residual, a constrained back propagation learning algorithm, which satisfies a Kuhn-Tucker inequality condition is proposed. Experimental results indicate that the prediction gain of the proposed NN predictor was not seriously decreased even when the constrained optimization algorithm was employed.

• Journal of Broadcast Engineering
• /
• v.20 no.2
• /
• pp.265-272
• /
• 2015

A new technology on video frame rate up-conversion (FRUC) is presented by combining the median filter and motion estimation (ME) with an occlusion detection (OD) method. First, ME is performed to have a motion vector. Then, the OD method is used to refine motion vector in the occlusion region. Since the wrong motion vector can be obtained with high possibility in the occluded area, a median filtering that less depends on the motion vector is applied to that area, and since the motion vector is continuous and robust in the non-occluded area, BDMC(Bi-Directional Motion Compensated interpolation) is applied to obtain interpolated image in that area. BDMC using the bi-directional motion vectors achieves good results when continuity and robustness of the motion vector is higher. Experimental results show that the proposed algorithm provides better performance than the conventional approach. The average gain of PSNR (Peak Signal to Noise Ratio) is approximately 0.16 dB in the test sequences compared with BDMC.

• Journal of Korea Multimedia Society
• /
• v.6 no.3
• /
• pp.508-517
• /
• 2003

This paper presents perceptual model with a stochastic multiresolution characteristic that can be applied with watermark embedding in the biorthogonal wavelet domain. The perceptual model with adaptive watermarking algorithm embeds at the texture and edge region for more strongly embedded watermark by the SSQ. The watermark embedding is based on the computation of a NVF that has local image properties. This method uses non- stationary Gaussian and stationary Generalized Gaussian models because watermark has noise properties. The particularities of embedding in the stationary GG model use shape parameter and variance of each subband regions in multiresolution. To estimate the shape parameter, we use a moment matching method. Non-stationary Gaussian model uses the local mean and variance of each subband. The experiment results of simulation were found to be excellent invisibility and robustness. Experiments of such distortion are executed by Stirmark 3.1 benchmark test.

• Journal of Korea Multimedia Society
• /
• v.15 no.11
• /
• pp.1284-1291
• /
• 2012

Ship object detection is a technique to detect the existence and the location of ship when ship objects are shown on input image sequence, and there are wide variations in accuracy due to environmental changes and noise of input image. In order to solve this problem, in this paper, we propose multiple ship object detection based on background registration technique and morphology operation. The proposed method consists of the following five steps: background elimination step, noise elimination step, object standard position setting step, object restructure step, and multiple object detection steps. The experimental results show accurate and real-time ship detection for 15 different test sequences with a detection rate of 98.7%, and robustness against variable environment. The proposed method may be helpful as the base technique of sea surface monitoring or automatic ship sailing.

• The Journal of the Acoustical Society of Korea
• /
• v.18 no.2
• /
• pp.11-17
• /
• 1999

This paper proposes an observation probability smoothing technique for the robustness of a discrete hidden Markov(DHMM) model based speech recognizer. Also, an appropriate noise robust processing in car environment is suggested from experimental results. The noisy speech is often mislabeled during the vector quantization process. To reduce the effects of such mislabelings, the proposed technique increases the observation probability of similar codewords. For the noise robust processing in car environment, the liftering on the distance measure of feature vectors, the high pass filtering, and the spectral subtraction methods are examined. Recognition experiments on the 14-isolated words consists of the Korean digits and command words were performed. The database was recorded in a stopping car and a running car environments. The recognition rates of the baseline recognizer were 97.4% in a stopping situation and 59.1% in a running situation. Using the proposed observation probability smoothing technique, the liftering, the high pass filtering, and the spectral subtraction the recognition rates were enhanced to 98.3% in a stopping situation and to 88.6% in a running situation.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.31 no.12C
• /
• pp.1184-1193
• /
• 2006

Image interpolation has been widely used and studied in the various fields of image processing. There are many approaches of varying complexity and robustness. In this paper, a new distance weight is proposed for the conventional linear interpolation. In comparison with the conventional linear weight, the new distance weight uses a quadratic or cubic polynomial equation to reflect that the interpolated value should be influenced more by the value of closer pixels in an input image. In this paper, the new adaptive linear (NAL) interpolation, which considers patterns near the interpolated value, is also proposed. This algorithm requires a pattern weight, which is used to determine the ratio of reflection on local patterns, to obtain an interpolated image that exhibits better quality at various magnification factors (MF). In the computer simulation, not only did the NAL interpolation exhibit much lower computational complexity than conventional bicubic interpolation, it also improved peak signal-to-noise ratios (PSNR).

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.36 no.1C
• /
• pp.48-56
• /
• 2011

Cognitive radio (CR) systems have received significant interest as a promising solution to the spectral shortage problem through efficient use of the frequency spectrum by opportunistically exploiting unlicensed frequency bands. Orthogonal frequency division multiplexing (OFDM) is widely regarded as a highly promising candidate for CR systems. However, the frequency bands used by CR systems are expected to suffer from non-Gaussian noise, which considerably degrades the performance of the conventional OFDM carrier frequency offset (CFO) estimation schemes. In this paper, robust CFO estimation schemes for OFDM-based CR systems in non-Gaussian channels are proposed. Simulation results demonstrate that the proposed estimators offer robustness and substantial performance improvement over the conventional estimator.

• Smart Structures and Systems
• /
• v.20 no.2
• /
• pp.127-137
• /
• 2017

Identification of damping characteristics is of significant importance for dynamic response analysis and condition assessment of structural systems. Damping is associated with the behavior of the energy dissipation mechanism. Identification of damping ratios based on the sensitivity of dynamic responses and the model updating technique is investigated with numerical and experimental investigations. The effectiveness and performance of using the sensitivity-based model updating method and vibration monitoring data for damping ratios identification are investigated. Numerical studies on a three-dimensional truss bridge model are conducted to verify the effectiveness of the proposed approach. Measurement noise effect and the initial finite element modelling errors are considered. The results demonstrate that the damping ratio identification with the proposed approach is not sensitive to the noise effect but could be affected significantly by the modelling errors. Experimental studies on a steel planar frame structure are conducted. The robustness and performance of the proposed damping identification approach are investigated with real measured vibration data. The results demonstrate that the proposed approach has a decent and reliable performance to identify the damping ratios.