• The Journal of the Acoustical Society of Korea
• /
• v.29 no.8
• /
• pp.509-515
• /
• 2010

This paper presents a new method for upmixing mono signal to stereo signal with guaranteeing high stereophonic image quality (SIQ) and large apparent source width (ASW). The proposed method consists of analysis phase and synthesis phase. In analysis phase, a mono signal is first decomposed into multiple sound sources by the use of high-rank nonnegative matrix factorization. Then the multiple sources are clustered into two groups based on tonality criterion. In synthesis phase, one group is directly fed into left and right channels while the other group is decorrelated before being fed into each channel. Subjective tests reveals that the proposed method gives listener high SIQ and large ASW with minimizing timbral distortions.

• The Journal of the Acoustical Society of Korea
• /
• v.34 no.3
• /
• pp.234-239
• /
• 2015

Non-Negative Matrix Factorization (NMF) is a method for updating dictionary and gain in alternating manner. Due to ease of implementation and intuitive interpretation, NMF is widely used to detect and separate overlapping sound events. However, NMF that utilizes non-negativity constraints generates parts-based representation and this distinct property leads to a dictionary containing fragmented acoustic events. As a result, the presence of shared basis results in performance degradation in both separation and detection tasks of overlapping sound events. In this paper, we propose a new method that utilizes K-Singular Value Decomposition (K-SVD) based dictionary to address and mitigate the part-based representation issue during the dictionary learning step. Subsequently, we calculate the gain using NMF in sound event detection step. We evaluate and confirm that overlapping sound event detection performance of the proposed method is better than the conventional method that utilizes NMF based dictionary.

• The Journal of Information Systems
• /
• v.29 no.3
• /
• pp.237-251
• /
• 2020

Purpose The research was studied the hierarchical Hangul emotion index by organizing all the emotions which SNS users are thinking. As a preliminary study by the researcher, the English-based Plutchick (1980)'s emotional standard was reinterpreted in Korean, and a hashtag with implicit meaning on SNS was studied. To build a multidimensional emotion dictionary and classify three-dimensional emotions, an emotion seed was selected for the composition of seven emotion sets, and an emotion word dictionary was constructed by collecting SNS hashtags derived from each emotion seed. We also want to explore the priority of each Hangul emotion index. Design/methodology/approach In the process of transforming the matrix through the vector process of words constituting the sentence, weights were extracted using TF-IDF (Term Frequency Inverse Document Frequency), and the dimension reduction technique of the matrix in the emotion set was NMF (Nonnegative Matrix Factorization) algorithm. The emotional dimension was solved by using the characteristic value of the emotional word. The cosine distance algorithm was used to measure the distance between vectors by measuring the similarity of emotion words in the emotion set. Findings Customer needs analysis is a force to read changes in emotions, and Korean emotion word research is the customer's needs. In addition, the ranking of the emotion words within the emotion set will be a special criterion for reading the depth of the emotion. The sentiment index study of this research believes that by providing companies with effective information for emotional marketing, new business opportunities will be expanded and valued. In addition, if the emotion dictionary is eventually connected to the emotional DNA of the product, it will be possible to define the "emotional DNA", which is a set of emotions that the product should have.

• ETRI Journal
• /
• v.34 no.6
• /
• pp.958-961
• /
• 2012

In this letter, we present a new algorithm for the harmonic and percussive separation of jazz music. Using a short-time Fourier transform and nonnegative matrix factorization, the signal is decomposed into rank components. Each component is then split into harmonic and percussive parts using masks calculated based on their tonalities. Finally, the harmonic and percussive parts are separated after applying the masks and a summation. We evaluate the algorithm based on real audio examples using both objective and subjective assessments. The proposed algorithm performs well for the separation of harmonic and percussive parts of jazz excerpts.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2014.06a
• /
• pp.9-11
• /
• 2014

본 논문에서는 오디오 신호의 특성 표현에 유용한 nonnegative matrix factorization(NMF)에 대해 설명하였으며, expectation maximization (EM)을 이용한 NMF 파라미터 추출 및 EM-NMF 기반한 오디오 소스 분리 기술에 대해서 설명했다. 또한, 다중 단계 NMF-EM 구조의 객체 분리를 통해서 객체 분리 성능을 향상시키기 위한 알고리즘을 제안하며, 제안된 알고리즘은 K-pop 음원과 SDR(source distortion ratio)를 통해서 객체 분리 성능을 평가한다. 성능 평가 결과 제안된 알고리즘은 다중 단계를 통해 약 3dB 의 보컬 분리 성능이 향상되며, 상업적 음원 제작에서 사용되는 가상 오디오 효과가 많이 적용된 음원에서 약 5dB 의 분리 성능을 향상시켰다. 그러므로 제안된 방식은 오디오 객체 분리에 유용한 방법이 될 것으로 생각된다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2014.06a
• /
• pp.5-8
• /
• 2014

본 논문에서는 비음수 행렬 분해 (NMF) 기법을 이용하여 단일 채널에서 배경음 성분을 추출하는 알고리즘에 대해 서술한다. 이러한 배경음 성분 추출은 오디오 업믹싱 시스템을 고려하여 개발되었으며, 기존의 연구를 통하여 분리된 배경음 신호가 서라운드 채널 혹은 상방향 채널에 적용될 경우 청취자의 공간감을 향상시킬 수 있다는 사실이 이미 확인된 바 있다. 다만 기존의 기법은 음향 신호를 모두 축적하여 일괄적으로 처리해야 한다는 단점이 있어, 스트리밍 시스템이나 디지털 신호 프로세서 등을 이용한 시스템에서 사용될 수 없는 단점이 있다. 본 논문에서는 이를 해소하기 위하여 실시간 비음수 행렬 분해 기법을 이용한 배경음 추출 시스템을 고안하여 실험하였다. 실험 결과 실시간 배경음 추출 기법이 신호의 후반부에서는 원하는 대로 동작하나, 초중반에 기저가 과도하게 설정되는 문제점이 있음을 확인할 수 있었으며, 이에 대한 해결이 향후 연구 과제가 될 것이다.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.41 no.8
• /
• pp.851-858
• /
• 2016

This paper presents a new statistical voice activity detection (VAD) based on the probabilistic interpretation of nonnegative matrix factorization (NMF). The objective function of the NMF using Kullback-Leibler divergence coincides with the negative log likelihood function of the data if the distribution of the data given the basis and encoding matrices is modeled as Poisson distributions. Based on this probabilistic NMF, the VAD is constructed using the likelihood ratio test assuming that speech and noise follow Poisson distributions. Experimental results show that the proposed approach outperformed the conventional Gaussian model-based and NMF-based methods at 0-15 dB signal-to-noise ratio simulation conditions.

• Journal of Information Science Theory and Practice
• /
• v.8 no.2
• /
• pp.6-17
• /
• 2020

Word similarity is often measured to enhance system performance in the information retrieval field and other related areas. This paper reports on an experimental comparison of values for word similarity measures that were computed based on 50 intentionally selected words from a Reuters corpus. There were three targets, including (1) co-occurrence-based similarity measures (for which a co-occurrence frequency is counted as the number of documents or sentences), (2) context-based distributional similarity measures obtained from a latent Dirichlet allocation (LDA), nonnegative matrix factorization (NMF), and Word2Vec algorithm, and (3) similarity measures computed from the tf-idf weights of each word according to a vector space model (VSM). Here, a Pearson correlation coefficient for a pair of VSM-based similarity measures and co-occurrence-based similarity measures according to the number of documents was highest. Group-average agglomerative hierarchical clustering was also applied to similarity matrices computed by individual measures. An evaluation of the cluster sets according to an answer set revealed that VSM- and LDA-based similarity measures performed best.

• Proceedings of the Korean Society for Bioinformatics Conference
• /
• 2001.10a
• /
• pp.165-196
• /
• 2001

DNA chip 또는 microarray는 다수의 유전자 또는 유전자 조각을 (보통 수천내지 수만 개)칩상에 고정시켜 놓고 DNA hybridization 반응을 이용하여 유전자들의 발현 양상을 분석할 수 있는 기술이다. 이러한 high-throughput기술은 예전에는 생각하지 못했던 여러가지 분자생물학의 문제에 대한 해답을 제시해 줄 수 있을 뿐 만 아니라, 분자수준에서의 질병 진단, 신약 개발, 환경 오염 문제의 해결 등 그 응용 가능성이 무한하다. 이 기술의 실용적인 적용을 위해서는 DNA chip을 제작하기 위한 하드웨어/웻웨어 기술 외에도 이러한 데이터로부터 최대한 유용하고 새로운 지식을 창출하기 위한 bioinformatics 기술이 핵심이라고 할 수 있다. 유전자 발현 패턴을 데이터마이닝하는 문제는 크게 clustering, classification, dependency analysis로 구분할 수 있으며 이러한 기술은 통계학과인공지능 기계학습에 기반을 두고 있다. 주로 사용된 기법으로는 principal component analysis, hierarchical clustering, k-means, self-organizing maps, decision trees, multilayer perceptron neural networks, association rules 등이다. 본 세미나에서는 이러한 기본적인 기계학습 기술 외에 최근에 연구되고 있는 새로운 학습 기술로서 probabilistic graphical model (PGM)을 소개하고 이를 DNA chip 데이터 분석에 응용하는 연구를 살펴본다. PGM은 인공신경망, 그래프 이론, 확률 이론이 결합되어 형성된 기계학습 모델로서 인간 두뇌의 기억과 학습 기작에 기반을 두고 있으며 다른 기계학습 모델과의 큰 차이점 중의 하나는 generative model이라는 것이다. 즉 일단 모델이 만들어지면 이것으로부터 새로운 데이터를 생성할 수 있는 능력이 있어서, 만들어진 모델을 검증하고 이로부터 새로운 사실을 추론해 낼 수 있어 biological data mining 문제에서와 같이 새로운 지식을 발견하는 exploratory analysis에 적합하다. 또한probabilistic graphical model은 기존의 신경망 모델과는 달리 deterministic한의사결정이 아니라 확률에 기반한 soft inference를 하고 학습된 모델로부터 관련된 요인들간의 인과관계(causal relationship) 또는 상호의존관계(dependency)를 분석하기에 적합한 장점이 있다. 군체적인 PGM 모델의 예로서, Bayesian network, nonnegative matrix factorization (NMF), generative topographic mapping (GTM)의 구조와 학습 및 추론알고리즘을소개하고 이를 DNA칩 데이터 분석 평가 대회인 CAMDA-2000과 CAMDA-2001에서 사용된cancer diagnosis 문제와 gene-drug dependency analysis 문제에 적용한 결과를 살펴본다.

• The Korean Journal of Nuclear Medicine
• /
• v.38 no.1
• /
• pp.21-29
• /
• 2004

Purpose: Much evidence suggests long-term cigarette smoking alters coronary vascular endothelial response. On this study, we applied nonnegative matrix factorization (NMF), an unsupervised learning algorithm, to CO-less $H_2^{15}O-PET$ to investigate coronary endothelial dysfunction caused by smoking noninvasively. Materials and methods: This study enrolled eighteen young male volunteers consisting of 9 smokers $(23.8{\pm}1.1\;yr;\;6.5{\pm}2.5$ pack-years) and 9 nonsmokers $(23.8{\pm}2.9 yr)$. They do not have any cardiovascular risk factor or disease history. Myocardial $H_2^{15}O-PET$ was performed at rest, during cold ($5^{\circ}C$) pressor stimulation and during adenosine infusion. Left ventricular blood pool and myocardium were segmented on dynamic PET data by NMF method. Myocardial blood flow (MBF) was calculated from input and tissue functions by a single compartmental model with correction of partial volume and spillover effects. Results: There were no significant difference in resting MBF between the two groups (Smokers: 1.43 0.41 ml/g/min and non-smokers: $1.37{\pm}0.41$ ml/g/min p=NS). during cold pressor stimulation, MBF in smokers was significantly lower than 4hat in non-smokers ($1.25{\pm}0.34$ ml/g/min vs $1.59{\pm}0.29$ ml/gmin; p=0.019). The difference in the ratio of cold pressor MBF to resting MBF between the two groups was also significant (p=0.024; $90{\pm}24%$ in smokers and $122{\pm}28%$ in non-smokers.). During adenosine infusion, however, hyperemic MBF did not differ significantly between smokers and non-smokers ($5.81{\pm}1.99$ ml/g/min vs $5.11{\pm}1.31$ ml/g/min ; p=NS). Conclusion: in smokers, MBF during cold pressor stimulation was significantly lower compared wi4h nonsmokers, reflecting smoking-Induced endothelial dysfunction. However, there was no significant difference in MBF during adenosine-induced hyperemia between the two groups.