• 방송공학회논문지
• /
• 제11권2호
• /
• pp.200-206
• /
• 2006

본 논문은 음성 부호화기에서 입력 오디오 신호가 보다 효과적으로 처리되도록 하기 위해 입력오디오신호를 전 처리하는 방법을 소개한다. 이를 위해 본 논문은 잡음억제 및 적응이득제어 방법을 도입한다. 여기서 입력 오디오 신호는 잡음 부가된 신호로 간주되며 그 오디오 신호의 부호화오차신호는 부가된 잡음신호로 간주된다. 입력 오디오 신호는 기존의 잡음억제방식에 따라 잡음신호 즉, 부호화 오차신호가 억제된 뒤 적응이득제어기를 거쳐 최종적으로 음성 부호화기에 인가된다. 결과적으로 이러한 동작을 통하여 입력 오디오 신호의 주파수 스펙트럼 분포가 음성 부호화기 특성에 맞게 재배치된다 이 방법의 하나의 단점은 부호화 오차를 계산하기 위해 사전에 추가적인 부호화 동작이 필요하다는 것이다. 반면, 이 방법은 일반적인 구조를 가지고 있으며 따라서 기존의 여러 음성부호화기에 쉽게 적용될 수 있다는 장점을 가진다. 주관적인 선호도 조사결과 제안된 방법이 복잡한 음악신호로 기인한 성가신 잡음을 사전에 억제해 주며 결과적으로 음질개선을 가져다준다는 것을 확인할 수 있었다.

• 방송공학회논문지
• /
• 제25권3호
• /
• pp.374-385
• /
• 2020

본 논문에서 Equirectangular Projection(ERP) 영상으로 행동 인식을 할 때의 문제점들을 해결할 수 있는 전처리 기법을 제안한다. 본 논문에서 제안하는 전처리 기법은 사람 객체를 행동의 주체 즉, Object of Interest(OOI)로 가정하고, OOI의 주변 영역을 ROI로 가정한다. 전처리 기법은 3개의 모듈로 이루어져 있다. I) 객체 인식 모델로 영상 내 사람 객체를 인식한다. II) 입력 영상에서 saliency map을 생성한다. III) 인식된 사람 객체와 saliency map을 이용하여 행동의 주체를 선정한다. 이후 행동 인식 모델에 선정된 행동의 주체 boundary box를 입력하여 행동 인식 성능을 높인다. 제안하는 전처리기법을 사용한 데이터를 행동 인식 모델에 입력한 방법의 성능과 원본 ERP 영상을 입력한 방법의 성능을 비교하였을 때 최대 99.6%의 성능 향상을 보이며, OOI가 감지되는 프레임만을 추출하였을 때 행동 관련 영상 요약의 효과도 볼 수 있다.

• 한국경영과학회:학술대회논문집
• /
• 한국경영과학회 1996년도 추계학술대회발표논문집; 고려대학교, 서울; 26 Oct. 1996
• /
• pp.285-288
• /
• 1996

Generally MPS, standardized by IBM, is the input type of large scale linear programming problems, and there may be unnecessary variables or constraints. These can be discarded by preprocessing. As the size of a problem is reduced by preprocessing, the running time is reduced. And more, the infeasibility of a problem may be detected before using solution methods. When the preprocessing implemented by this paper is used in NETLIB problems, it removes unnecessary variables and constraints by 21%, 15%, respectively. The use of preprocessing gives in the average 21% reduction in running time by applying the interior point method. Preprocessing can detect 10 out of 30 infeasible NETLIB problems.

• 제어로봇시스템학회논문지
• /
• 제3권6호
• /
• pp.626-631
• /
• 1997

This paper represents the design of a real-time image preprocessing system. The preprocessing system performs hardware-wise mask operations and thresholding operations at the speed of camera output single rate. The preprocessing system consists of the preprocessing board and the main processing board. The preprocessing board includes preprocessing unit that includes a $5\times5$ mask processor and LUT, and can perform mask and threshold operations in real-time. To achieve high-resolution image input data($20485\timesn$), the preprocessing board has a linescan camera interface. The main processing board includes the image processor unit and main processor unit. The image processor unit is equipped with TI's TMS320C32 DSP and can perform image processing algorithms at high speed. The main processor unit controls the operation of total system. The proposed system is faster than the conventional CPU based system.

• 한국경영과학회:학술대회논문집
• /
• 한국경영과학회 1997년도 추계학술대회발표논문집; 홍익대학교, 서울; 1 Nov. 1997
• /
• pp.105-108
• /
• 1997

Previous studies in stock market predictions using artificial intelligence techniques such as artificial neural networks and case-based reasoning, have focused mainly on spot market prediction. Korea launched trading in index futures market (KOSPI 200) on May 3, 1996, then more people became attracted to this market. Thus, this research intends to predict the daily up/down fluctuant direction of the price for KOSPI 200 index futures to meet this recent surge of interest. The forecasting methodologies employed in this research are the integration of genetic algorithm and artificial neural network (GAANN) and the integration of genetic algorithm and case-based reasoning (GACBR). Genetic algorithm was mainly used to select relevant input variables. This study adopts the categorical data preprocessing based on expert's knowledge as well as traditional data preprocessing. The experimental results of each forecasting method with each data preprocessing method are compared and statistically tested. Artificial neural network and case-based reasoning methods with best performance are integrated. Out-of-the Model Integration and In-Model Integration are presented as the integration methodology. The research outcomes are as follows; First, genetic algorithms are useful and effective method to select input variables for Al techniques. Second, the results of the experiment with categorical data preprocessing significantly outperform that with traditional data preprocessing in forecasting up/down fluctuant direction of index futures price. Third, the integration of genetic algorithm and case-based reasoning (GACBR) outperforms the integration of genetic algorithm and artificial neural network (GAANN). Forth, the integration of genetic algorithm, case-based reasoning and artificial neural network (GAANN-GACBR, GACBRNN and GANNCBR) provide worse results than GACBR.

• Journal of Information Processing Systems
• /
• 제17권2호
• /
• pp.399-410
• /
• 2021

Aiming at the problem that complex external variables in natural scenes have a greater impact on facial expression recognition results, a facial expression recognition method based on two-stream convolutional neural network is proposed. The model introduces exponentially enhanced shared input weights before each level of convolution input, and uses soft attention mechanism modules on the space-time features of the combination of static and dynamic streams. This enables the network to autonomously find areas that are more relevant to the expression category and pay more attention to these areas. Through these means, the information of irrelevant interference areas is suppressed. In order to solve the problem of poor local robustness caused by lighting and expression changes, this paper also performs lighting preprocessing with the lighting preprocessing chain algorithm to eliminate most of the lighting effects. Experimental results on AFEW6.0 and Multi-PIE datasets show that the recognition rates of this method are 95.05% and 61.40%, respectively, which are better than other comparison methods.

• 전기학회논문지
• /
• 제58권1호
• /
• pp.173-180
• /
• 2009

It is difficult to predict non-stationary or chaotic time series which includes the drift and/or the non-linearity as well as uncertainty. To solve it, we propose an effective prediction method which adopts data preprocessing and multiple model TS fuzzy predictors combined with model selection mechanism. In data preprocessing procedure, the candidates of the optimal difference interval are determined based on the correlation analysis, and corresponding difference data sets are generated in order to use them as predictor input instead of the original ones because the difference data can stabilize the statistical characteristics of those time series and better reveals their implicit properties. Then, TS fuzzy predictors are constructed for multiple model bank, where k-means clustering algorithm is used for fuzzy partition of input space, and the least squares method is applied to parameter identification of fuzzy rules. Among the predictors in the model bank, the one which best minimizes the performance index is selected, and it is used for prediction thereafter. Finally, the error compensation procedure based on correlation analysis is added to improve the prediction accuracy. Some computer simulations are performed to verify the effectiveness of the proposed method.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2005년도 ICCAS
• /
• pp.2485-2489
• /
• 2005

This paper introduces a new preprocessing scheme to replace original data of gray scale images with particular ordered data so that performance of lossless compression can be improved more efficiently. As a kind of preprocessing technique to maximize performance of entropy encoder, the proposed method converts the input image data into more compressible form. Before encoding a stream of the input image, the proposed preprocessor counts co-occurrence frequencies for neighboring pixel pairs. Then, it replaces each pair of adjacent gray values with particular ordered numbers based on the investigated co-occurrence frequencies. When compressing ordered image using entropy encoder, we can expect to raise compression rate more highly because of enhanced statistical feature of the input image. In this paper, we show that lossless compression rate increased by up to 37.85% when comparing results from compressing preprocessed and non-preprocessed image data using entropy encoder such as Huffman, Arithmetic encoder.

• 음성과학
• /
• 제10권4호
• /
• pp.125-136
• /
• 2003

Speech coding techniques have been studied to reduce the complexity and bit rate but also to improve the sound quality. CELP type vocoder, has used as a one of standard, supports the great sound quality even low bit rate. In this paper, the preprocessing of input speech to reduce the bit rate is the different with the conventional vocoder. The different kinds of parameter are used for the preprocessing so this paper is compared with theses parameters for finding the more appropriate parameter for the vocoder. The parameters are used to synthesize the speech not to encode or decode for coding technique so we proposed the simple algorithm not to have the influence on the processing time or the computation time. The parameters in used the preprocessing step are speaking rate, duration and PSOLA technique.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2001년도 하계종합학술대회 논문집(4)
• /
• pp.93-96
• /
• 2001

In this paper, we present a vision algorithm and method for input image improvement and preprocessing of dented and raised characters on the sidewall of tires. we define optical condition between reflect coefficient and reflectance by the physical vector calculate. On the contrary this work will recognize the engraved characters using the computer vision technique. Tire input images have all most same grey levels between the characters and backgrounds. The reflectance is little from a tire surface. therefore, it's very difficult segment the characters from the background. Moreover, one side of the character string is raised and the other is dented. So, the captured images are varied with the angle of camera and illumination. For optimum input images, the angle between camera and illumination was found out to be with in 90。 .In addition, We used complex filtering with low-pass and high-pass band filters to improve input images, for clear input images. Finally we define equation reflect coefficient and reflectance. By doing this, we obtained good images of tires for pattern recognition.