• The Journal of the Acoustical Society of Korea
• /
• v.14 no.1
• /
• pp.31-39
• /
• 1995

In this paper, we have constructed phoneme discriminant filters [PDF] according to the linear discriminant function. These discriminant filters do not follow the heuristic rules by the experts but the mathematical methods in iterative learning. Proposed system. is based on the piecewise linear classifier and error correction learning method. The segmentation of speech and the classification of phoneme are carried out simutaneously by the PDF. Because each of them operates independently, some speech intervals may have multiple outputs. Therefore, we introduce the unified coefficients by the output unification process. But sometimes the output has a region which shows no response, or insensitive. So we propose time windows and median filters to remove such problems. We have trained this system with the 549 monosyllables uttered 3 times by 3 male speakers. After we detect the endpoint of speech signal using threshold value and zero crossing rate, the vowels and consonants are separated by the PDF, and then selected phoneme passes through the following PDF. Finally this system unifies the outputs for competitive region or insensitive area using time window and median filter.

• The Transactions of the Korea Information Processing Society
• /
• v.3 no.5
• /
• pp.1307-1316
• /
• 1996

In this paper, we propose a vector quantization method which uses a dynamic address mapping based on exploring the high interblock correlation. In the proposed method, we reduce bit-rate by defining an address transform function, which maps a VQ address of an input block which will be encoded into a new address in the reordered codebook by using side match error. In one case that an original address can be transformed into a new transformed address which is lower than the threshold value, we encode the new address of the transformed convector, and in the other case we encode the address of the original convector which is not transformed. Experimental results indicate that the proposed scheme reduces the bit-rate by 45~50% compared with the ordi-nary VQ method forimage compression, at the same quality of the reconstructed image as that of the ordinary VQ system.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.14 no.7
• /
• pp.677-684
• /
• 2003

This paper addresses the subblock phase weighting(SPW) method to reduce the PAPR in OFDM system. This method divides the input block of OFDM signal into many subblocks and lower the peak power by weighting the phase of each subblocks properly. SPW method can be realized by only one IFFT. PAPR reduction performance is novelly examined when the adjacent, interleaved and random subblock partitioning schemes are used in the SPW system. The random subblock partition scheme has the most effective. More subblocks can effectively reduce the PAPR, but there is a problem that the processing time of iteration is increased. We propose a new weighting factor combination of the complementary sequence characteristic with threshold technique. OFDM data can be recovered by the inserted side information of weighting factor in the feed forward type. Also, BER performance of this SPW system is analyzed when error happens in the side information.

• Journal of Broadcast Engineering
• /
• v.18 no.4
• /
• pp.515-524
• /
• 2013

In this paper, we propose a novel face authentication method based on LBP matching of individual major blocks in mobile environment. In order to construct individual major blocks from photos, we find the blocks that have the highest similarity and use different numbers of blocks depending on the probability distribution by applying threshold. And, we use multi-radius LBP histograms in the determination of individual major blocks to improve performance of generic LBP histogram based approach. By using the multi-radius LBP histograms in face authentication, we can successfully reduce the false acceptance rate compare to the previous methods. Also, we can see that the proposed method shows low error rate about 7.72% compare to the pervious method in spite of use small number of blocks about 44.59% only.

• Journal of Korean Society for Geospatial Information Science
• /
• v.13 no.3 s.33
• /
• pp.25-32
• /
• 2005

This paper presents a method for point-based 3D building reconstruction using Lidar data and digital map. The proposed method consists of three processes: extraction of building roof points, identification of roof types, and 3D building reconstruction. After extracting points inside the polygon of building, the ground surface, wall and tree points among the extracted points are removed through the filtering process. The filtered points are then fitted into the flat plane using ODR(Orthogonal Distance Regression) in the first place. If the fitting error is within the predefined threshold, the surface is classified as a flat roof. Otherwise, the surface is fitted and classified into a gable or arch roof through RMSE analysis. Experimental results showed that the proposed method classified successfully three different types of roof and that the fusion of LIDAR data and digital map could be a feasible method of modeling 3D building reconstruction.

• Journal of the Korean Society of Radiology
• /
• v.8 no.7
• /
• pp.435-441
• /
• 2014

This paper suggests techniques to enhance coding time which is a problem in traditional fractal compression and to improve fidelity of reconstructed images by determining fractal coefficient through adaptive selection of block approximation formula. First, to reduce coding time, we construct a linear list of domain blocks of which characteristics is given by their luminance and variance and then we control block searching time according to the first permissible threshold value. Next, when employing three-level block partition, if a range block of minimum partition level cannot find a domain block which has a satisfying approximation error, There applied to 24-bpp color image compression and image techniques. The result did not occur a loss in the image quality of the image when using the encoding method, such as almost to the color in the RGB image compression rate and image quality, such as gray-level images and showed good.

• The Journal of the Acoustical Society of Korea
• /
• v.20 no.3
• /
• pp.56-62
• /
• 2001

In this paper, we adopt a missing data theory to speech recognition. It can be used in order to maintain high performance of speech recognizer when the missing data occurs. In general, hidden Markov model (HMM) is used as a stochastic classifier for speech recognition task. Acoustic events are represented by continuous probability density function in continuous density HMM(CDHMM). The missing data theory has an advantage that can be easily applicable to this CDHMM. A marginalization method is used for processing missing data because it has small complexity and is easy to apply to automatic speech recognition (ASR). Also, a spectral subtraction is used for detecting missing data. If the difference between the energy of speech and that of background noise is below given threshold value, we determine that missing has occurred. We propose a new method that examines the reliability of detected missing data using voicing probability. The voicing probability is used to find voiced frames. It is used to process the missing data in voiced region that has more redundant information than consonants. The experimental results showed that our method improves performance than baseline system that uses spectral subtraction method only. In 452 words isolated word recognition experiment, the proposed method using the voicing probability reduced the average word error rate by 12％ in a typical noise situation.

• Journal of the Korean Society for Nondestructive Testing
• /
• v.21 no.3
• /
• pp.281-287
• /
• 2001

For the conventional two-dimensional source location of acoustic emission (AE) based on the threshold crossing, wave velocity has to be measured in the actual structure to calculate the arrival-time difference and thus to form the two hyperbolae. Velocity is dependent on the fiber orientation, however, due to the dependence of elastic modulus on fiber orientation in anisotropic materials such as compost#e plates. This tan affect the accuracy of AE source location and make the source location procedure complicated. In this study, we propose a method to reduce the location error in anisotropic plates by using the numerical solution of nonlinear equations, where the velocity term has been removed by employing the fourth sensor. The efficiency and validity of the proposed method has also been experimentally verified.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.43 no.6 s.312
• /
• pp.76-82
• /
• 2006

In this paper, the biomedical signal compression method, which is combined with the multi-stage vector quantization and wavelet lifting scheme, is proposed. It utilizes the property of wavelet coefficients that give emphasis on approximation coefficients. The transmitted codebook index consists of the code vectors obtained by wavelet lifting coefficients of ECG and error signals from the 1024 block length, respectively. Each codebook is adaptively updated by the method comparing to the distance of input codevectors with candidate codevectors by using an pre-defined threshold value. The proposed compression method showed blow 3% in term of PRD and 276.62 bits/sec in term of CDR.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.17 no.3
• /
• pp.705-711
• /
• 2013

QRS detection of ECG is the most popular and easy way to detect cardiac-disease. But it is difficult to analyze the ECG signal because of various noise types. Also in the healthcare system that must continuously monitor people's situation, it is necessary to process ECG signal in realtime. In other words, the design of algorithm that exactly detects QRS wave using minimal computation and classifies PVC by analyzing the persons's physical condition and/or environment is needed. Thus, efficient QRS detection and PVC classification based on profiling method is presented in this paper. For this purpose, we detected QRS through the preprocessing method using morphological filter, adaptive threshold, and window. Also, we applied profiling method to classify each patient's normal cardiac behavior through hash function. The performance of R wave detection, normal beat and PVC classification is evaluated by using MIT-BIH arrhythmia database. The achieved scores indicate the average of 99.77% in R wave detection and the rate of 0.65% in normal beat classification error and 93.29% in PVC classification.