• The Journal of the Acoustical Society of Korea
• /
• v.38 no.1
• /
• pp.47-50
• /
• 2019

This paper proposes an acoustic model adaptation method for effective speech recognition in noisy environments. In the proposed algorithm, the noise corruption function is estimated employing DNN (Deep Neural Network), and the function is applied to the model parameter estimation. The experimental results using the Aurora 2.0 framework and database demonstrate that the proposed model adaptation method shows more effective in known and unknown noisy environments compared to the conventional methods. In particular, the experiments of the unknown environments show 15.87 % of relative improvement in the average of WER (Word Error Rate).

• Transactions of the Korean Society of Mechanical Engineers B
• /
• v.27 no.6
• /
• pp.726-735
• /
• 2003

A Process of 3-D Particle image velocimetry, called here, as '3-D volume PIV' was developed for the full-field measurement of 3-D complex flows. The present method includes the coordinate transformation from image to camera, calibration of camera by a calibrator based on the collinear equation, stereo matching of particles by the approximation of the epipolar lines, accurate calculation of 3-D particle positions, identification of velocity vectors by 3-D cross-correlation equation, removal of error vectors by a statistical method followed by a continuity equation criterior, and finally 3-D animation as the post processing. In principle, as two frame images only are necessary for the single instantaneous analysis 3-D flow field, more effective vectors are obtainable contrary to the previous multi-frame vector algorithm. An Experimental system was also used for the application of the proposed method. Three analog CCD camera and a Halogen lamp illumination were adopted to capture the wake flow behind a bluff obstacle. Among 200 effective particle s in two consecutive frames, 170 vectors were obtained averagely in the present study.

• Transactions of the Korean Society of Mechanical Engineers B
• /
• v.27 no.6
• /
• pp.736-743
• /
• 2003

A process of 3-D particle image velocimetry, called here, as '3-D stereo PIV' was developed for the measurement of an illuminated slied section field of 3-D complex flows. The present method includes modeling of camera by a calibrator based on the homogeneous coordinate system, transfromation of the oblique-angled image to the right-angled image, identification of 2-D velocity vectors by 2-D cross-correlation equation, stereo matching of 2-D velocity vectors of two cameras, accurate calculation of 3-D velocity vectors by homogeneous coordinate system, removal of error vectors by a statistical method followed by a continuity equation criterior, and finally 3-D animation as the post processing. An experimental system was also used for the application of the proposed method. Three analog CCD cameras and an Argon-Ion Laser(300mW) for illumination were adopted to capture the wake flow behind a bluff obstacle.

• Journal of the Korea Convergence Society
• /
• v.9 no.2
• /
• pp.13-18
• /
• 2018

This paper is based on a simple approach that the most efficient learning of a multi-layered network is the process of finding the optimal set of weight vectors. To overcome the disadvantages of general learning problems, the proposed model uses a combination of features of EBP and OVSSA. In other words, the proposed method can construct a single model by taking advantage of each algorithm so that it can escape to the probability theory of OVSSA in order to reinforce the property that EBP falls into local minimum value. In the proposed algorithm, methods for reducing errors in EBP are used as energy functions and the energy is minimized to OVSSA. A simple experimental result confirms that two algorithms with different properties can be combined.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.9 no.5
• /
• pp.490-495
• /
• 1999

In this paper. a learning method VoD-EBP for neural networks is proposed, which learn patterns by error back propagation. Based on Voronoi diagram, the method initializes the weights of the neural networks systematically, wh~ch results in faster learning speed and alleviated local optimum problem. The method also shows better the reliability of the design of neural network because proper number of hidden nodes are determined from the analysis of Voronoi diagram. For testing the performance, this paper shows the results of solving the XOR problem and the parity problem. The results were showed faster learning speed than ordinary error back propagation algorithm. In solving the problem, local optimum problems have not been observed.

• 한국가시화정보학회:학술대회논문집
• /
• 2002.11a
• /
• pp.19-22
• /
• 2002

A process of 3-D particle image velocimetry, called here, as '3-D stereo PIV' was developed for the measurement of a section field of 3-D complex flows. The present method includes modeling of camera by a calibrator based on the homogeneous coordinate system, transfromation of oblique-angled image to transformed image, identification of 2-D velocity vectors by 2-D cross-correlation equation, stereo matching of 2-D velocity vectors of two cameras, accurate calculation of 3-D velocity vectors by homogeneous coordinate system and finally 3-D animation as the post processing. In principle, as two frame images only are necessary for the single instantaneous analysis of a section field of 3-D flow, more effective vectors are obtainable contrary to the previous multi-frame vector algorithm. An experimental system was also used for the application of the proposed method. Three analog CCD cameras and an Argon-Ion Laser(300mW) for illumination were adopted to capture the wake flow behind a bluff obstacle.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.42 no.3 s.303
• /
• pp.81-90
• /
• 2005

When we transmit signals, which are quantized by the vector quantizer (VQ), through noisy channels, the overall performance of the coding system is very dependent on the employed quantization scheme and the channel error effect. In order to design an optimal coding system, the source and channel coding scheme should be jointly optimized as in the channel-optimized VQ. As a suboptimal approach, we may consider the robust VQ (RVQ). In RVQ, we consider developing an index assignment function for mapping the output of quantizers to channel symbols so that the effect of the channel errors is minimized. Recently, a VQ, which can reduce the encoding complexity and is called the sample-adaptive product quantizer (SAPQ), has been proposed. SAPQ has very similar quantizer structure as to the product quantizer (PQ). However, the quantization performance can be better than PQ. Further, the encoding complexity and the memory requirement for the codebooks are lower than the regular full-search VQ case. In this paper, SAPQ is employed in order to design an RVQ to channel errors by reducing the vector dimension. Discussions on the codebook structure of SAPQ and experiments are introduced in an aspect of robustness to noisy channels.

• Journal of the Korea Society of Computer and Information
• /
• v.18 no.6
• /
• pp.21-28
• /
• 2013

This paper proposes an effective fire detection approach that includes the following multiple heterogeneous algorithms: moving region detection using grey level histograms, color segmentation using fuzzy c-means clustering (FCM), feature extraction using a grey level co-occurrence matrix (GLCM), and fire classification using support vector machine (SVM). The proposed approach determines the optimal threshold values based on grey level histograms in order to detect moving regions, and then performs color segmentation in the CIE LAB color space by applying the FCM. These steps help to specify candidate regions of fire. We then extract features of fire using the GLCM and these features are used as inputs of SVM to classify fire or non-fire. We evaluate the proposed approach by comparing it with two state-of-the-art fire detection algorithms in terms of the fire detection rate (or percentages of true positive, PTP) and the false fire detection rate (or percentages of true negative, PTN). Experimental results indicated that the proposed approach outperformed conventional fire detection algorithms by yielding 97.94% for PTP and 4.63% for PTN, respectively.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.52 no.8
• /
• pp.117-125
• /
• 2015

In this paper, we propose a classification algorithm based on the obtained EEG(Electroencephalogram) signal for the control of 'left' and 'right' turnings of which a driving system composed of EEG sensor, Labview, DAQ, Matlab and driving robot. The proposed algorithm uses features extracted from frequency band information obtained by DWT (Discrete Wavelet Transform) and selects features of high discrimination by using Fisher score. We, also propose the number of feature vectors for the best classification performance by using SVM(Support Vector Machine) classifier and propose a decision pending algorithm based on MLD (Maximum Likelihood Decision) to prevent malfunction due to misclassification. The selected four feature vectors for the proposed algorithm are the mean of absolute value of voltage and the standard deviation of d5(2-4Hz) and d2(16-32Hz) frequency bands of P8 channel according to the international standard electrode placement method. By using the SVM classifier, we obtained 98.75% accuracy and 1.25% error rate. Also, when we specify error probability of 70% for decision pending, we obtained 95.63% accuracy and 0% error rate by using the proposed decision pending algorithm.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.10
• /
• pp.133-153
• /
• 2023

In this paper, we propose a novel approach to investigating brain-signal measurement technology using Electroencephalography (EEG). Traditionally, researchers have combined EEG signals with bio-signals (BSs) to enhance the classification performance of emotional states. Our objective was to explore the synergistic effects of coupling EEG and BSs, and determine whether the combination of EEG+BS improves the classification accuracy of emotional states compared to using EEG alone or combining EEG with pseudo-random signals (PS) generated arbitrarily by random generators. Employing four feature extraction methods, we examined four combinations: EEG alone, EG+BS, EEG+BS+PS, and EEG+PS, utilizing data from two widely-used open datasets. Emotional states (task versus rest states) were classified using Support Vector Machine (SVM) and Long Short-Term Memory (LSTM) classifiers. Our results revealed that when using the highest accuracy SVM-FFT, the average error rates of EEG+BS were 4.7% and 6.5% higher than those of EEG+PS and EEG alone, respectively. We also conducted a thorough analysis of EEG+BS by combining numerous PSs. The error rate of EEG+BS+PS displayed a V-shaped curve, initially decreasing due to the deep double descent phenomenon, followed by an increase attributed to the curse of dimensionality. Consequently, our findings suggest that the combination of EEG+BS may not always yield promising classification performance.