잡음 변동에 강인한 MAP 진폭 스펙트럼 추정기를 이용한 음성강조 기법

김학윤(Hack-Yoon Kim) 한국정보기술학회 2016 한국정보기술학회논문지 Vol.14 No.2

The conventional MAP(maximum a posteriori) estimation technology is used to the separation of speech signals from noisy. In noise range of this algorithm, the observed signal is assumed to be non-speech from the initial interval signals. It can be determined from its region of spectrum and also used as reference spectrum. If an input noise signal is changed, this noise environment cannot be managed because of problems of the above technique. In this paper, a new spectral amplitude estimator for acoustical background noise suppression based on MAP estimation and weighted noise estimation is proposed. Since the spectral amplitude estimator cannot suppress sinusoidal signals generated from arbitrary noise variation of an observed signal, an adaptive line enhancer(ALE) to the spectral amplitude estimator for eliminating the sinusoids is also added. The proposed spectral estimator algorithm with ALE is verified by using computer simulation.

형태 기반 밝기 보정 모델을 이용한 감귤의 결점 검출 알고리즘

김학윤(Hack-Yoon Kim),윤종민(Jong-Min Yoon) 한국정보기술학회 2016 한국정보기술학회논문지 Vol.14 No.4

Detection of defects on fruits utilized the lightness information correction of the acquired image of fruit in previous algorithm. This method is correcting the dark edge of lightness image of fruit and detecting the fruit defect to a threshold value. However, this method not only had a high possibility of false detection according to the size and location of defect as well as its limitation in fruit’s shape. In this paper, the lightness correction model is created using by the shape of information of mandarin and we have proposed an method to detect fruit’s defects. The performance was assessed by comparing the proposed method and conventional method. Experimental results shows that the lightness correction is feasible regardless of the existing problem of conventional method which are the size and location of defects also shape of mandarin. It is also verified through the comparison of the proposed method and improved by 10% of recognition rate.

DMB를 위한 3차원 오디오 신호처리 기술

김학윤(Hack-Yoon Kim) 한국정보기술학회 2010 한국정보기술학회논문지 Vol.8 No.5

As the 3D Audio Reproduction algorithm is the technology that restores the lost spatial cue, which was lost given recording and playback, it obtains information same as the form in sound location, environment, and a sound source, which occurred by granting the spatial distribution information to auditory sense. This conventional 3D Audio Reproduction algorithm is placed too emphasis on the specific frequency band due to the gain value for stressing the stereo image, thereby having a problem that sharp noise components take place. Accordingly, in this paper suggested the new 3D Audio Reproduction algorithm by using Gammatone auditory filter in order to solve the problem about the existing algorithm. As a result of comparing, SRS was produced the sphere that a signal is excessively amplified more than the original sound. The proposed algorithm was swelled the size of the signal while having the character similar to the original sound, but there was the excessive amplification phenomenon, which was produced in SRS.

머리 움직임을 고려한 3차원 음 재생 방법에 관한 연구

김학윤(Hack-Yoon Kim) 한국정보기술학회 2014 한국정보기술학회논문지 Vol.12 No.2

As the 3D Sound reproduction algorithm is the technology that restores the lost spatial cue, which was lost given recording and playback, it obtains information same as the form in sound location, environment, and a sound source, which occurred by granting the spatial distribution information to auditory sense. Many conventional methods are proposed for implementing the 3D sound reproduction system. These methods have the problem which is not focused and considered on an effect of head movement. Hence, This paper develops and implements the adaptive 3D system with head movement sensor to improve previous ones. Performance of the proposed system is evaluated through listening tests. It shows the better results of the method for headphone than speaker and proves the better experimental results of the head movement than those of excluding the head movement.

한국어 음성 명료도를 이용한 커스텀 디지털 보청기의 피팅 알고리즘에 관한 연구

김학윤(Hack-Yoon Kim) 한국정보기술학회 2020 한국정보기술학회논문지 Vol.18 No.1

The currently used major algorithms for fitting insertion gain of digital hearing aids to hearing impairment in Korea stem from an overseas development. Among conventional fitting algorithms, the NAL-NL2 algorithm was developed with an emphasis on articulation of speech. Since the algorithm was developed based on the articulation of English, it is considered inappropriate to apply it to hearing impairments in Korea. Therefore, this study proposes the Revised NAL-NL2(RNAL-NL2) algorithm suitable for hearing patients in Korea by embedding Korean articulation based on NAL-NL2 algorithm to compensate for these defects. In addition, the validity of the proposed algorithm is verified by comparative analysis of relationship between insertion gain characteristics and input/output, and then this algorithm is applied to actual hearing impairments to verify the usefulness of the proposed algorithm.

한국어 명료도를 이용한 난청 분류 기법에 관한 연구

김학윤(Hack-Yoon Kim),오형택(Hyeong-Taek Oh),한철수(Cheolsu Han),유동구(Dong-Gu Yoo) 한국정보기술학회 2016 Proceedings of KIIT Conference Vol.2016 No.6

스펙트럼 감산을 이용한 잡음 억제 방법의 음질 개선

김학윤(Kim Hack-Yoon) 한국정보기술학회 2007 한국정보기술학회논문지 Vol.5 No.3

In this paper, some methods, which improve speech quality and suppress musical noise in noise canceller based on spectral subtraction approach, have been discussed. First, in order to suppress musical noise, the spectrum is weighted so as to suppress the high frequency components, which are dominant in the musical noise. Second, in order to improve voice quality, the original noisy voice is added to the voice after noise cancellation. these methods have been simulated and compared to the conventional methods. The simulation results show the proposed methods can improve the voice quality and suppress the musical noise. However, the method adding the original noisy voice requires more proper threshold control.

모폴로지 특성 분석 방법을 이용한 충돌성 잡음 제거 기법

김학윤(Hack-Yoon Kim) 한국정보기술학회 2018 한국정보기술학회논문지 Vol.16 No.1

In this paper, we propose a separation method using Morphological Component Analysis(MCA) for mixtures which consist of speeches and impact noises. We assume that the speech is stationary within a short period of time and the duration of the noises is shorter than the period. MCA is a separation method for signal mixtures using a pair of dictionary which can yield a sparse representation for signal but not appropriate for other. However, it is difficult for impact noises to define specific base because the only assumption of noises if that its durations are relatively short. Then, we represented impact noises with subspaces that consists of signals whose durations are restricted. Speeches are represented with DFT bases. Signals are separated by imposing a cost function of sparsity on coefficients of DFT and choice of subspace for noises. In a computer based simulation, we demonstrate the noise reduction capability of the proposed method for various impact noises.

OATSP를 이용한 마이크로폰의 주파수 특성 응답 측정 알고리즘

박병욱,김학윤,Park, Byoung-Uk,Kim, Hack-Yoon 한국음향학회 2007 韓國音響學會誌 Vol.26 No.2

마이크로폰의 주파수 응답 특성은 마이크로폰이 레벨 허용 범위로 재생할 수 있는 주파수 범위를 나타내는 것으로, 마이크로폰이 가지고 있는 특성을 평가하는 기준으로 사용되는 가장 중요한 음향 특성 파라메타 중의 하나이다. 이와 같은 마이크로폰의 주파수 응답 특성을 측정하기 위한 기존의 방법들은 그 측정 조건이 매우 까다로울 뿐만 아니라, 고가의 장비를 사용하여 측정하여야 한다는 문제점을 갖고 있다. 이러한 단점을 보완하기 위하여 본 논문에서는 마이크로폰의 주파수 응답 특성을 간단하게 측정할 수 있는 알고리즘을 제안한다. 제안한 알고리즘은 컴퓨터로 생성한 Optimized Aoshima's Time Stretched Pulse(OATSP) 신호를 표준 스피커를 통하여 발생시킨 다음, 측정하고자 하는 마이크로폰으로 수음된 신호와 역 OATSP 신호를 컨볼루션시켜 마이크로폰의 임펄스 응답을 측정하고, 이 신호를 이용하여 측정할 마이크로폰의 주파수 응답 특성을 구하는 방범이다. 제안한 알고리즘의 성능 평가는 제안한 알고리즘을 이용하여 구한 마이크로폰의 주파수 응답 특성 측정값과 그들이 갖고 있던 주파수 응답 특성 데이터를 비교 분석하였다. 비교 결과, 측정한 각각의 마이크로폰 주파수 응답 특성들 사이에 오차가 발생하였으나, 오차가 그 측정값들이 허용 오차(${\pm}3{\sim}{\pm}5dB$) 범위에 내에 있었으므로 제안한 알고리즘이 마이크로폰의 주파수 응답 특성을 측정하기에 적합한 방법임을 입증하였다. The frequency response of a microphone, which indicates the frequency range that a microphone can output within the approved level, is one of the most significant standards used to measure the characteristics of a microphone. At present, conventional methods of measuring the frequency response are complicated and involve the use of expensive equipment. To complement the disadvantages, this paper suggests a new algorithm that can measure the frequency response of a microphone in a simple manner. The algorithm suggested in this paper generates the Optimized Aoshima's Time Stretched Pulse(OATSP) signal from a computer via a standard speaker and measures the impulse response of a microphone by convolution the inverse OATSP signal and the received by the microphone to be measured. Then, the frequency response of the microphone to be measured is calculated using the signals. The performance test for the algorithm suggested in the study was conducted through a comparative analysis of the frequency response data and the measures of frequency response of the microphone measured by the algorithm. It proved that the algorithm is suitable for measuring the frequency response of a microphone, and that despite a few errors they are all within the error tolerance.

비파괴 선별시스템을 위한 당도 추정 모델의 성능 비교

윤종민(Jong-Min Yoon),소민혁(Min-Hyeok So),한철수(Cheolsu Han),김학윤(Hack-Yoon Kim) 한국정보기술학회 2014 Proceedings of KIIT Conference Vol.2014 No.5

근적외선을 이용한 사과 선별시스템은 사과를 파괴하지 않고 신속하게 선별할 수 있기 때문에 근래 많은 사과 선별장에서 이용되고 있다. 그러나 선별장의 실측 환경을 고려한 사과 당도 추정 모델은 검토되고 있지 않는 실정이다 따라서 본 연구에서는 선별장의 사과 선별시스템에서 실측된 흡광도 데이터를 이용하여 다중 회귀 부분 최소제곱회귀 능형회귀 모델을 각각 설계하고 그 정밀도를 비교 평가한다.