• Proceedings of the Safety Management and Science Conference
• /
• 2005.11a
• /
• pp.300-310
• /
• 2005

Mahalanobis Taguchi System (MTS) is developed by Genishi Taguchi as a part of his quality engineering methodology. The basic idea of Taguchi's quality engineering is looking for the way of effectiveness of analyzing multivariate system. In the MTS, with the standardized variables of healthy normal data, Mahalanobis Distance(MD) calculated and that can be discriminate between normal and abnormal objects. If this discrimination process is successful, next step is optimization which is try to reduce number of attributes by neglecting less effective attributes to MD. Orthogonal Array (OA) and Signal to Noise ratio (S/N) are used to evaluate the amount contribution of each attribute to the MD. Wisconsin Breast Cancer study, from machining learning repository at University of California at Irvine, used for examining the discriminant ability of MTS.

• Proceedings of the KIEE Conference
• /
• 2000.11d
• /
• pp.761-764
• /
• 2000

This paper presents a Fuzzy Neural Network (FNN) system to decide whether or not the right information of lanes can be extracted from gray-level images of road scene. The decision of noisy level of input images has been required because much noises usually deteriorates the performance of feature detection based on image processing and lead to erroneous results. As input parameters to FNN, eight noisiness indexes are constructed from a cumulative distribution function (CDF) and proved the indexes being classifiers of images as the good and the bad corrupted by sources of noise by correlation analysis between input images and the indexes. Considering real-time processing and discrimination efficiency, the proposed FNN is structured by eight input parameters, three fuzzy variables and single output. We conduct much experiments and show that our system has comparable performance in terms of false-positive rates.

• Proceedings of the KIEE Conference
• /
• 2001.11c
• /
• pp.459-460
• /
• 2001

In this paper, we analyze algorithms for diagnosing of VEP(visual evoked potential) signal. We used wavelet transform for the preprocessing of VEP signal data and back propagation neural network for the pattern recognition. We used several wavelets to study their effects and efficiency in the preprocessing of VEP. The diagnosis system led to good results. We obtained the noise reduced and compressed signal with the wavelet transform of the training VEP signal. So it is possible to train the neural network faster and exact diagnosis processing is possible in the neural network. From the experimental results, we know that the discrimination ability of the neural network is changed by the type of basis vector and the proposed system is good to the diagnosis of VEP.

• 제어로봇시스템학회:학술대회논문집
• /
• 1995.10a
• /
• pp.303-306
• /
• 1995

The present paper deals with extraction of figures and characters from their background using the knowledge of color. At each pixel of the image on the CRT sent from a video camera, RGB values are transformed into the values in another color system, HSI, where "H" denotes hue;"S" denotes saturation;"I" denotes intensity. Representing color in HSI color space is advantageous, since a human feels color mainly in hue with the aid of brightness and purity. Comparing HSI data thus obtained with the masked original image detects noise-free edges included in the orginal image. Then setting a set of HSI thresholds and changing it identifies the portion of image of the same color. This color information is used in recongnizing characters and figures as an auxiliary system of a hierachical figure categorization method for characters and figures recognition.cters and figures recognition.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.14 no.6
• /
• pp.699-712
• /
• 1989

This paper describes a method of Korean speech synhes is using format VOCODER. The parameters of speech synthes is are a follows, 1) format F1, F2, and F3 by spectrum moment method and F4, F5 using the length of vocal tract. 2) pitch frequencies obtained by optimu, Comb method using AMDF. 3) short time average energy and short time mean amplitude. 4) The decision method of bandwidth reportd by Fant. 5) voicde/unvoiced discrimination using zerocrossing. 6) excitation wave reported by Rosenberg. 7) gaussian white noise. Synthesis results are in fairly good agreement with original speech.

• Journal of the Korean Institute of Telematics and Electronics A
• /
• v.31A no.6
• /
• pp.61-69
• /
• 1994

For distortion-invariant recognition of Korean characters, a holographic implementation of an optical associative memory system is proposed. The structure of the proposed system is a single-layer neural network employing interconneclion matrix, thresholding and feedback. To provide the interconnection matrix, we use two CGII's which are placed on intermcdiate plane of cascaded Vander Lugt corrclators to form an optical memory loop. The holographic correlator stores reference images in a hologram and retrives them in a coherently illuminated feedback loop. An input image which maybe noisy or incomplete, is applicd to the system and simultaneously correlated optically with all of the stord images. These correlations are throsholed and fed back to the input, where the strongest correlation reinforces the input image. The enhanced image passes arround the loop repeatedly, approaching the stored image more closely on each pass until the system stabilizes on the desired image. The computer simulation results show that the proposed Korean Character recognition algorithm has high discrimination capability and noise immunity.

• Proceedings of the IEEK Conference
• /
• 2006.06a
• /
• pp.283-284
• /
• 2006

In this paper, we propose a novel segmentation and classification method using texture features for the document image. First, we extract the local entropy and then segment the document image to separate the background and the foreground using the Otsu's method. Finally, we classify the segmented regions into each component using PCA(principle component analysis) algorithm based on the texture features that are extracted from the co-occurrence matrix for the entropy image. The entropy-based segmentation is robust to not only noise and the change of light, but also skew and rotation. Texture features are not restricted from any form of the document image and have a superior discrimination for each component. In addition, PCA algorithm used for the classifier can classify the components more robustly than neural network.

• Proceedings of the KIEE Conference
• /
• 2006.07c
• /
• pp.1469-1471
• /
• 2006

It is very important to discriminate noises such as air corona in measuring on-site partial discharges (PD). Therefore, in order to investigate the possibility of separating PD(partial discharge) and noises through the pulse analysis, pulse shapes measured by a resistive sensor and a HFCT were analyzed and PA(pulse analysis) were performed. For the purpose, the HFPD detection and PA system has been developed. Also void discharges and air corona were adopted as the artificial defect and noises, respectively.

• Proceedings of the KIEE Conference
• /
• 2006.07c
• /
• pp.1535-1537
• /
• 2006

Recently, the noise discrimination method using PD pulse waveshape analysis has been suggested to be very effective method which can improve the reliability of the on-site PD measurement. In this method, the data clusters due to PD pulses or noises can be distinguished on the PA map. And for the automatic recognition of the PD clusters, it is necessary to adopt the adaptable pattern recognition method. In this study, as for the algorithm which can recognize data clusters, the data mining method has been adopted and the result of the analysis has been reported.

• Proceedings of the KSPS conference
• /
• 2005.11a
• /
• pp.59-62
• /
• 2005

Voice activity detection (VAD) is important in many areas of speech processing technology. Speech/nonspeech discrimination in noisy environments is a difficult task because the feature parameters used for the VAD are sensitive to the surrounding environments. Thus the VAD performance is severely degraded at low signal-to-noise ratios (SNRs). In this paper, a new VAD algorithm is proposed based on the degree of voicing and Quantile SNR (QSNR). These two feature parameters are more robust than other features such as energy and spectral entropy in noisy environments. The effectiveness of proposed algorithm is evaluated under the diverse noisy environments in the Aurora2 DB. According to out experiment, the proposed VAD outperforms the ETSI Advanced Frontend VAD.