• The Journal of the Acoustical Society of Korea
• /
• v.35 no.4
• /
• pp.303-309
• /
• 2016

This paper proposes a novel approach to the problem of audio signal clustering using a stacked autoencoder. The proposed stacked autoencoder learns an efficient representation for the input signal, enables clustering constituent signals with similar characteristics, and therefore the original sources can be separated based on the clustering results. STFT (Short-Time Fourier Transform) is performed to extract time-frequency spectrum, and rectangular windows at all the possible locations are used as input values to the autoencoder. The outputs at the middle, encoding layer, are used to cluster the rectangular windows and the original sources are separated by the Wiener filters derived from the clustering results. Source separation experiments were carried out in comparison to the conventional NMF (Non-negative Matrix Factorization), and the estimated sources by the proposed method well represent the characteristics of the orignal sources as shown in the time-frequency representation.

• The Journal of the Acoustical Society of Korea
• /
• v.12 no.2E
• /
• pp.33-37
• /
• 1993

부호여기선형예측 형태의 음성 부호화기에 있어서주된 단점은 요구되는 계산량이많다는 것이다. 본 논문에서는 음질의 저하없이 자기상관에 근거한 피치예측기에서 피치 찾는 시간을 줄이는 간단한 방법을 제안한다. 기본적인 발상은, 피치찾는 과정에서 나타나는 자기상관 함수의 특성에 근거하여, 음의 봉우리의 폭을 이전 양의 봉우리의 폭을 이용하여 추정하여 건너 뜀으로써 피치 찾는 범위를 자기상관함수의 양의 값쪽으로 제한하는 것이다. 이렇게 피치찾는 구간에 제약을 가함으로써 요구되는 계산량은 감소하게 된다. 제안된 방법을 적용한 결과, 평균 약 40%의 계산량 감소를 음질의 저하없이 얻을 수 있다.

• Journal of Korea Multimedia Society
• /
• v.11 no.11
• /
• pp.1480-1490
• /
• 2008

Design of a linear predictor and matching of an entropy coder is the art of lossless audio coding. In this paper, we use the covariance method and the Choleskey decomposition for calculating linear prediction coefficients instead of the autocorreation method and the Levinson-Durbin recursion. These results are compared to the polynomial predictor. Both of them, the predictor which has small prediction error is selected. For the entropy coding, we use the Golomb-Rice coder using the block-based parameter estimation method and the sequential adaptation method with LOCO-land RLGR. The proposed predictor and the block-based parameter estimation have $2.2879%{\sim}0.3413%$ improved compression ratios compared to FLAC lossless audio coder which use the autocorrelation method and the Levinson-Durbin recursion. The proposed predictor and the LOCO-I adaptation method could improved by $2.2879%{\sim}0.3413%$. But the proposed predictor and the RLGR adaptation method got better results with specific signals.

• Proceedings of the Acoustical Society of Korea Conference
• /
• autumn
• /
• pp.119-122
• /
• 2004

G.723.1은 저 전송률 환경에서 고 음질을 제공하여 주고 있으나 CELP형 부호화기가 갖는 합성에 의한 분석(Analysis by Synthesis)방식의 구조로 인해 많은 처리 시간과 계산량을 요구하게 된다. 본 논문에서는 G.723.1에 대해 NAMDF함수를 적용하여 델타 피치 검색과정의 계산량을 줄여 부호화기의 전체 계산량을 감소시키는 방법을 제안하였다. 기존의 피치 검출 알고리즘에서 피치 검출을 위해 사용하고 있는 자기상관함수는 곱셈 연산에서 발생하는 bit의 dynamic range가 커서 나눗셈 연산에서도 과도한 연산량을 필요로 한다. 따라서, 이러한 계산량의 감소를 위해 기존의 자기상관함수 대신 계산량을 감소하기 위하여 NAMDF 방법을 적용하였고 추가된 skipping 기법을 사용하였다. 계산량 감소율 측면에서는 약 $64\%$의 감소율을 보였고 기존의 방법과 제안한 방법간의 피치 pitch contour은 원음성의 피치 contour와 유사하였고, 음질 평가에서도 기존의 G.723.1 부호화기 합성음과 유사한 길과를 얻을 수 있었다.

• Annual Conference on Human and Language Technology
• /
• 2022.10a
• /
• pp.624-629
• /
• 2022

에세이의 창의성을 자동으로 분류하는 기존의 주요 연구는 말뭉치에서 빈번하게 등장하지 않는 단어에 초점을 맞추어 기계학습을 수행한다. 그러나 이러한 연구는 에세이의 주제와 상관없이 단순히 참신한 단어가 많아 창의적으로 분류되는 문제점이 발생한다. 본 논문에서는 어텐션(Attention)과 역문서 빈도(Inverse Document Frequency; IDF)를 이용하여 에세이 내용 전달에 있어 중요하면서 참신한 단어에 높은 가중치를 두는 문맥 벡터를 구하고, 자기부호화기(AutoEncoder) 모델을 사용하여 문맥 벡터들로부터 창의적인 에세이와 창의적이지 않은 에세이의 특징 벡터를 추출한다. 그리고 시험 단계에서 새로운 에세이의 특징 벡터와 비교하여 그 에세이가 창의적인지 아닌지 분류하는 딥러닝 모델을 제안한다. 실험 결과에 따르면 제안 방안은 기존 방안에 비해 높은 정확도를 보인다. 구체적으로 제안 방안의 평균 정확도는 92%였고 기존의 주요 방안보다 9%의 정확도 향상을 보였다.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2003.10a
• /
• pp.122-125
• /
• 2003

An adaptive contour smoothing algorithm designed as a preprocessor for shape coders is presented. In the proposed method, the degree of the adaptive smoothing is controlled based on the significance of each contour point, which is quantified according to inter-region contrast in an intensity image. The actual smoothing consists of an expansion operator and a thinning algorithm. Experimental results show that the proposed method results in a saving of about 20% in number of coded bits with a negligible additional texture degradation in the reconstructed intensity image.

• Journal of Advanced Navigation Technology
• /
• v.4 no.1
• /
• pp.67-77
• /
• 2000

This paper proposes a reduction scheme for codebook search time in the adaptive codebook using wavelet transformed coefficients. In a CELP coder, pitch estimation with a combined open loop and closed loop search in adaptive codebook needs a lengthy search. More precisely, the pitch search using autocorrelation function over all possible ranges has been shown inefficient compared to the consuming time. In this paper, we propose a new adaptive codebook search algorithm which ensures the same position for the pitch with maximum wavelet coefficient over various scaling factors in Dyadic wavelet transform. A new adaptive codebook search algorithm reduces 25% conventional search time with almost the same quality of speech.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.19 no.8
• /
• pp.1587-1594
• /
• 1994

The pitch predictor is one of the most important part for the robust tree coder. The hybrid backward pitch adapation which is a combination of a block adaptation and a recursive adaptation is used for the pitch predictor. In order to improve the error performance and track the pitch period change of the input speech, it is proposed to smooth the input of the pitch predictor. The smoother with three taps can have fixed coefficients or variable coefficients depending on the estimated autocorrelation function of the output of the pitch synthesizer. The inclusion of a variable smoother can track the pitch period change within a block and reduce the effect of channel errors.

• The Journal of the Acoustical Society of Korea
• /
• v.13 no.3
• /
• pp.33-40
• /
• 1994

Code Excited Linear Prediction(CELP) speech coders exhibit good performance at data rates below 4.8 kbps. This major drawback of CELP type coders is required much computation. In this paper, we propose a new pitch search method that preserves the quality of the CELP vocoder with reducing complexity. In the pitch searching, we detect the segments of high correlation by a simple preprocessing, and then carry out the pitch searching only for the segments obtained by the preprocessing. By using the proposed method, we can get approximately $77\%$ complexity reduction in the pitch search.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.13 no.7
• /
• pp.3188-3193
• /
• 2012

It is important to search a pitch for vocoder. The major drawback to vocoders is their large computational requirements in searching a pitch and a codebook. In this paper, a simple method is proposed to improve the pitch searching process in the pitch filter almost without degradation of quality. The period of speech signal is emphasized by using Dual Pulse technique, the same type of autocorrelation method, in pitch search. Sometimes the incorrect pitch can be obtained by halving, doubling and trifling, To solve it, before searching a pitch, we estimate energy rate in a frame and compensate envelop of signal with it. By using the proposed algorithm in pitch search, its required computation are reduced and searching pitch is improved.