• Proceedings of the Korean Information Science Society Conference
• /
• 2006.06a
• /
• pp.391-393
• /
• 2006

인터넷의 급속한 확산과 대량 정보의 이동은 문서의 요약을 더욱 필요로 하고 있다. 본 논문은 비음수 행렬 인수분해로(NMF, non-negative matrix factorization) 얻어진 비음수 의미 가변 행렬(NSVM, non-negative semantic variable matrix)을 이용하여 자동으로 포괄적 문서요약 하는 새로운 방범을 제안하였다. 제안된 방법은 인간의 인식 과정과 유사한 비음수 제약을 사용한다. 이 결과 잠재의미색인에 비해 더욱 의미 있는 문장을 선택하여 문서를 요약할 수 있다. 또한, 비지도 학습에 의한 문서요약으로 사전 전문가에 의한 학습문장이 필요 없으며, 적은 계산비용을 통하여 쉽게 문장을 추출할 수 있는 장점을 갖는다.

• 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.

• Journal of KIISE:Software and Applications
• /
• v.31 no.10
• /
• pp.1412-1417
• /
• 2004

Sparse coding or independent component analysis (ICA) which is a holistic representation, was successfully applied to elucidate early auditor${\gamma}$ processing and to the task of sound classification. In contrast, parts-based representation is an alternative way o) understanding object recognition in brain. In this thesis we employ the non-negative matrix factorization (NMF) which learns parts-based representation in the task of sound classification. Methods of feature extraction from the spectro-temporal sounds using the NMF in the absence or presence of noise, are explained. Experimental results show that NMF-based features improve the performance of sound classification over ICA-based features.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2013.05a
• /
• pp.1651-1652
• /
• 2013

• The Journal of the Korea Contents Association
• /
• v.9 no.4
• /
• pp.121-130
• /
• 2009

In this paper, feature extraction methods, which is one field of reducing dimensions of high-dimensional data, are empirically investigated. We selected the traditional PCA(Principal Component Analysis), ICA(Independent Component Analysis), NMF(Non-negative Matrix Factorization), and sNMF(Sparse NMF) for comparisons. ICA has a similar feature with the simple cell of V1. NMF implemented a "parts-based representation in the brain" and sNMF is a improved version of NMF. In order to visually investigate the extracted features, handwritten digits are handled. Also, the extracted features are used to train multi-layer perceptrons for recognition test. The characteristic of each feature extraction method will be useful when applying feature extraction methods to many real-world problems.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.41 no.6
• /
• pp.175-186
• /
• 2004

We present a hybrid approach of texture-based method and connected component (CC)-based method for text extraction in complex images. Two primary methods, which are mainly utilized in this area, are sequentially merged for compensating for their weak points. An automatically constructed MLP-based texture classifier can increase recall rates for complex images with small amount of user intervention and without explicit feature extraction. CC-based filtering based on the shape information using NMF enhances the precision rate without affecting overall performance. As a result, a combination of texture and CC-based methods leads to not only robust but also efficient text extraction. We also enhance the processing speed by adopting appropriate region marking methods for each input image category.

• The Journal of the Acoustical Society of Korea
• /
• v.31 no.5
• /
• pp.332-339
• /
• 2012

Beamspace transform algorithm transforms spatial-domain data - such as x, y, z dimension - into incidence-angle-domain data, which is called beamspace-domain data. The beamspace transform method is generally used in source localization and tracking, and adaptive beamforming problem. When the beamspace transform method is used in multichannel audio source separation, the inverse beamspace transform is also important because the source image have to be reconstructed. This paper studies the beamspace transform and inverse transform algorithms for multichannel audio source separation system, especially for the beamspace-domain multichannel NMF algorithm.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.49 no.2
• /
• pp.53-60
• /
• 2012

In this paper, we proposes a new document summarization method using the expanded query by wikipedia and the semantic feature representing inherent structure of document set. The proposed method can expand the query from user's initial query using the relevance feedback based on wikipedia in order to reflect the user require. It can well represent the inherent structure of documents using the semantic feature by the non-negative matrix factorization (NMF). In addition, it can reduce the semantic gap between the user require and the result of document summarization to extract the meaningful sentences using the expanded query and semantic features. The experimental results demonstrate that the proposed method achieves better performance than the other methods to summary document.

• The Journal of the Korea Contents Association
• /
• v.18 no.1
• /
• pp.173-184
• /
• 2018

In human walking, muscle co-contraction which produces simultaneous activities of multiple muscles is important in motor control mechanism of the central nervous system. This study aims to understand positive and negative covariation mechanism of inter-muscle activities during walking. In this study, we measured electromyography (EMG) in leg muscles. To identify motor modules, we recored EMG from 4 leg muscles bilaterally (the tibialis anterior, medial gastrocnemius, rectus femoris and medial hamstring muscles) and performed non-negative matrix factorization (NMF) and principa component analysis (PCA). Then, we computed covariation values from various combinations between muscles or motor modules and used two-way repeated measures analysis of variance to identify significantly different covariation patterns between muscle combinations. As the results, we found significant differences between covariation values of muscle combinations (p < 0.05). muscle groups within the same motor modules produced the positive covariations. However, there were strong negative covariation between motor modules. There was negative covariation in all muscle combination. Stable inter-module negative covariation suggests that motor modules may be the control unit in the complex motor coordination.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.42 no.4 s.304
• /
• pp.79-86
• /
• 2005

In this paper, we propose a new LINF(Linear Independent Non-negative Factorization) algorithm for real-time face recognition systea This system greatly consists of the two parts: 1) face extraction part; 2) face recognition part. In the face extraction Part we applied subtraction image, the detection of eye and mouth region , and normalization method, and then in the face recognition Part we used LINF in extracted face candidate region images. The existing recognition system using only PCA(Principal Component Analysis) showed low recognition rates, and it was hard in the recognition system using only LDA(Linear Discriminants Analysis) to apply LDA directly when the training set is small. To overcome these shortcomings, we reduced dimension as the matrix that had non-negative value to be different from former eigenfaces and then applied LDA to the matrix in the proposed system We have experimented using self-organized DAIJFace database and ORL database offered by AT(')T laboratory in Cambridge, U.K. to evaluate the performance of the proposed system. The experimental results showed that the proposed method outperformed PCA, LDA, ICA(Independent Component Analysis) and PLMA(PCA-based LDA mixture algorithm) method within the framework of recognition accuracy.