• Proceedings of the IEEK Conference
• /
• 2001.09a
• /
• pp.583-586
• /
• 2001

본 논문에서는 다중 대역통과 표본화 이론에 기반하여 1.9GHz IS-95신호와 2.2GHz IMT-2000 신호를 하나의 ADC(analog to digital converter)를 사용하여 동시에 표본화하고 디지털 처리를 수행하는 방법을 제안하고 실험을 통하여 검증하였다. 일반적인 방법으로는 본 논문에서 고려하는 두 신호를 동시에 표본화하기 위해서는 표본화 주파수가 최소한 1GHz 이상의 고속의 ADC를 사용해야 한다. 그러나 현재 ADC의 기술은 광대역의 신호를 직접 더지털화하기에는 아직 미흡하다. 반면에 대역통과 표본화 이론은 기존의 상용 ADC와 기콘의 RF 시스템을 이용하여 다른 대역에 위치한 두 신호를 통합처리 할 수 있는 기반을 제공하고 있다. 본 논문에서는 이러한 대역통과 표본화 이론에 기반을 두고 상용 ADC를 사용하여 표본화 시스템을 구현하여 IS-95신호와 IMT-2000 신호를 표본화하고 이를 컴퓨터에서 디지털 필터를 이용하여 두 신호를 분리하는 실험을 통하여 다중 대역통과 표본화의 적용 가능성을 검증하였다.

• The Journal of the Acoustical Society of Korea
• /
• v.16 no.6
• /
• pp.110-114
• /
• 1997

When applying the nonuniform sampling to noisy speech signal, the required data rate increases to be comparable to or more than that by uniform sampling such as PCM. To solve this problem, we have proposed the waveform coding method, multiband nonuniform waveform coding(MNWC), applying the nonuniform sampling to band-separated speech signal[7]. However, the speech quality is deteriorated when it is compared to the uniform sampling method, since the high band is simply modeled as a Gaussian noise with average level. In this paper, as a good method to overcome this drawback, the high band is modeled as one of 16 codewords having different center frequencies. By doing this, with maintaining high speech quality as MOS score of average 3.16, the proposed method achieves 1.5 times higher compression ratio than that of the conventional nonuniform sampling method(CNSM).

• Journal of Internet Computing and Services
• /
• v.6 no.1
• /
• pp.105-114
• /
• 2005

This paper describes a new measuring technique by analysing multiple time-series patterns. This paper's goal is that extracts a really measured value having a sample pattern which is the best matched with an inputted time-series, and calculates a difference ratio with the value. Therefore, the proposed technique is not a recognition but a measurement. and not a hardware but a software. The proposed technique is consisted of three stages, initialization, learning and measurement. In the initialization stage, it decides weights of all parameters using importance given by an operator. In the learning stage, it classifies sample patterns using LBG and DTW algorithm, and then creates code sequences for all the patterns. In the measurement stage, it creates a code sequence for an inputted time-series pattern, finds samples having the same code sequence by hashing, and then selects the best matched sample. Finally it outputs the really measured value with the sample and the difference ratio. For the purpose of performance evaluation, we tested on multiple time-series patterns obtained from etching machine which is a semiconductor manufacturing.

• The KIPS Transactions:PartC
• /
• v.12C no.1 s.97
• /
• pp.63-68
• /
• 2005

In this paper, performance of a maximum-likelihood modulation classification for quadrature amplitude modulation (QAM) is studied. Unlike previous works, the relative classification performance with respect to the available modulations and performance limit with single-sample observation are presented. For those purposes, all constellations are set to have the same minimum Euclidean distance between symbols so that a smaller constellation is a subset of the larger ones. And only one sample of received waveform is used for multiple hypothesis test. As a result, classification performance is improved with increase in signal-to-noise ratio in all the experiments. Especially, when the true modulation format used in the transmitter is 4 QAM, almost perfect classification can be achieved without any additional information or observation samples. Though the possibility of false classification due to the symbols shared by subset constellations always exists, correct classification ratio of $80{\%}$ can be obtained with the single-sample observation when the true modulation formats are 16 and 64 QAM.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2012.10a
• /
• pp.69-72
• /
• 2012

In the bandpass sampling(BPS), the sampling frequency is lower than the frequency of the RF(radio frequency) signal being sampled. In this method, the baseband spectrum directly appears by the sampling itself, so that it is not necessary to use any down converter, making the receiver's hardware simpler. The second-order BPS uses two identical BPS samplers operating with an offset timing to each other. By a processing with their two sampled signals, it can be possible to cancel the aliasing or interference component if any due to the bandpass sampling. The interpolant filter, which is to manipulate the phase characteristics of the sampled signal, affects the performance of the cancellation. In this paper, a multiband interpolant filter is introduced, with which multiple interference signals from multiple RF bands can be cancelled simultaneously. We suggest several phase characteristics for the interpolant filter and have evaluated their performances through computer simulations. It has been shown that the filter with a continuous phase function gives the better performance.

• Journal of KIISE:Software and Applications
• /
• v.31 no.12
• /
• pp.1684-1690
• /
• 2004

This paper presents a method for classifier selection that uses distribution information of the training samples in a small region surrounding a sample. The conventional DCS-LA(Dynamic Classifier Selection - Local Accuracy) selects a classifier dynamically by comparing the local accuracy of each classifier at the test time, which inevitably requires long classification time. On the other hand, in the proposed approach, the best classifier in a local region is stored in the FSA(Feature Space Attribute) table during the training time, and the test is done by just referring to the table. Therefore, this approach enables fast classification because classification is not needed during test. Two feature space attributes are used entropy and density of k training samples around each sample. Each sample in the feature space is mapped into a point in the attribute space made by two attributes. The attribute space is divided into regular rectangular cells in which the local accuracy of each classifier is appended. The cells with associated local accuracy comprise the FSA table. During test, when a test sample is applied, the cell to which the test sample belongs is determined first by calculating the two attributes, and then, the most accurate classifier is chosen from the FSA table. To show the effectiveness of the proposed algorithm, it is compared with the conventional DCS -LA using the Elena database. The experiments show that the accuracy of the proposed algorithm is almost same as DCS-LA, but the classification time is about four times faster than that.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.50 no.7
• /
• pp.225-230
• /
• 2013

In a bandpass sampling (BPS), the frequency of the sampler is lower than that of the signal being sampled. In this method, the baseband spectrum directly appears by the sampling operation, so that it is not necessary to use any frequency down-converter, which makes the receiver's hardware simpler. The second-order BPS uses two identical BPS samplers, of which sampling times are offset by each other. By exploiting the relationship between two sampled signals, it can be possible to cancel the aliased signal component or the interference due to the bandpass sampling. In order to cancel the interference, an interpolant filter is used to manipulate the phase characteristics of the BPS sampled signal. In this paper, it is introduced a multiband interpolant filter which can simultaneously cancel multiple interference signals that have been aliased from multiple frequency bands. In case of no need of interference cancellation, another method is suggested to enhance the signal quality by 3dB. A computer simulation has been performed, and the feasibility of the suggested methods has been verified.

• Proceedings of the Korean Information Science Society Conference
• /
• 2000.10b
• /
• pp.431-433
• /
• 2000

본 논문에서는 얼굴인식에 주로 사용되는 PCA를 얼굴포즈판별로 적용해 보았다. 얼굴포즈판별은 개개인의 얼굴특징을 강조해야 하는 얼굴인식과는 달리 일반적인 얼굴특징을 이용하기 때문에 PCA에 적합한 응용분야이다. 그러나, 다양한 얼굴포즈에 대한 영상을 하나의 표본집합으로 사용하면, 표본집합의 분산이 크기 때문에 포즈별로 표본집합을 달리하여 PCA모듈을 구성하는 것이 타당하다. 표본수집의 어려움은 3차원 한국인 표준모형을 이용해 극복하고, 이를 통하여 다양한 조명방향 및 얼굴포즈에 대한 표본을 수집하였다. 5방향의 얼굴포즈에 대한 판별 실험을 통하여 모율화된 PCA의 분류기로서의 가능성을 살펴보고, 조명에 따른 오류를 완하하고자 비 선형적 패턴을 나타내는 각 PCA모듈의 결과를 신경망에 적용하여 보았다.

• Journal of the Korean Data and Information Science Society
• /
• v.11 no.1
• /
• pp.83-90
• /
• 2000

The chi-squared test statistic is usually employed for testing independence of two categorical variables in a two-way contingency table. It is well known that, under independence, the test statistic has an asymptotic chi-squared distribution under multinomial or product-multinomial models. For the case where both margins fixed, the sampling model of the contingency table is a multiple hypergeometric distribution and the chi-squared test statistic follows the same limiting distribution. In this paper, we study the difference between the small sample and large sample distributions of the chi-squared test statistic for the case with fixed margins. For a few small sample cases, the exact small sample distribution of the test statistic is directly computed. For a few large sample sizes, the small sample distribution of the statistic is generated via a Monte Carlo algorithm, and then is compared with the large sample distribution via chi-squared probability plots and Kolmogorov-Smirnov tests.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.48 no.5
• /
• pp.16-22
• /
• 2011

This paper presents the multi-object tracking approach using the background difference and particle filtering by monte carlo sampling. We apply particle filters based on probabilistic importance sampling to multi-object independently. We formulate the object observation model by the histogram distribution using color information and the object dynaminc model for the object motion information. Our approach does not increase computational complexity and derive stable performance. We implement the whole Bayesian maximum likelihood framework and describes robust methods coping with the real-world object tracking situation by the observation and transition model.