Article
Acoustics
Mathieu Fontaine, Kouhei Sekiguchi, Aditya Arie Nugraha, Yoshiaki Bando, Kazuyoshi Yoshii
Summary: This paper introduces a heavy-tailed extension method for blind source separation called GSM-FastMNMF, based on a wider class of heavy-tailed distributions. It outperforms existing methods in speech enhancement and separation.
IEEE-ACM TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING
(2022)
Article
Automation & Control Systems
Zhongqiang Luo, Chaofu Jing, Yan Chen, Xingzhong Xiong
Summary: This paper proposes a new anti-collision algorithm (MCV_NMF algorithm) for underdetermined RFID systems. The algorithm enhances performance by combining the independent principle of tag signals with nonnegative matrix factorization mechanism, effectively addressing the underdetermined collision problem in RFID systems.
Article
Engineering, Electrical & Electronic
Ondrej Mokry, Paul Magron, Thomas Oberlin, Cedric Fevotte
Summary: In this paper, a probabilistic framework based on nonnegative matrix factorization (NMF) is proposed to restore missing audio signal samples. Two expectation-maximization algorithms are derived to estimate the model parameters by treating the missing samples as latent variables, depending on the problem formulation in the time or time-frequency domain. Furthermore, an alternating minimization scheme is derived to address the novel problem by treating the missing samples as parameters. Experiments demonstrate the great convergence properties and competitive performance of the proposed methods in restoring short-to middle-length gaps in music signals.
Article
Computer Science, Information Systems
Mahbanou Zohrevandi, Saeed Setayeshi, Azam Rabiee, Midia Reshadi
Summary: This paper proposes a method for separating underdetermined convolutive blind speech in a multi-speaker environment based on mask prediction in the time-frequency domain. Clustering with a weighting function is used to consider parts of masks with potentially only one active source, and sparse filters are utilized in the time-domain to improve signal quality. Performance evaluation shows that the proposed method is more accurate than conventional algorithms.
MULTIMEDIA TOOLS AND APPLICATIONS
(2021)
Article
Computer Science, Artificial Intelligence
Haonan Huang, Guoxu Zhou, Qibin Zhao, Lifang He, Shengli Xie
Summary: In this article, a novel model called deep autoencoder-like NMF for MRL (DANMF-MRL) is proposed, which considers both multiview consistency and complementarity for a more comprehensive representation. A one-step DANMF-MRL is further proposed, which learns the latent representation and final clustering labels matrix in a unified framework, achieving optimal clustering performance without tedious clustering steps. Two efficient iterative optimization algorithms are developed with theoretical convergence analysis. Extensive experiments on five benchmark datasets demonstrate the superiority of the proposed approaches over other state-of-the-art MRL methods.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Engineering, Electrical & Electronic
Ping Yang, Ting-Zhu Huang, Jie Huang, Jin-Ju Wang
Summary: The proposed method incorporates weighted nuclear norm and $L_{1/2}$ norm to consider the low-rank and sparse priors of each abundance map simultaneously in hyperspectral unmixing. An adaptive update mechanism is implemented to treat each constraint differently, leading to improved unmixing effect and solving speed.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2022)
Article
Engineering, Biomedical
Yuan Xie, Kan Xie, Qiyu Yang, Shengli Xie
Summary: A novel blind source separation (BSS) method based on nonnegative matrix factorization (NMF) and auxiliary function technique is proposed for separating heart sound mixtures and lung sound mixtures in a reverberation environment. The method utilizes denoising pre-processing and source model optimization, and experimental results show its superiority over traditional methods in highly reverberant environments.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2021)
Article
Computer Science, Artificial Intelligence
Jin-Ju Wang, Ding-Cheng Wang, Ting-Zhu Huang, Jie Huang, Xi-Le Zhao, Liang-Jian Deng
Summary: In this paper, a new NTF-based model called EIC-NTF is proposed for hyperspectral unmixing to mitigate the impact of high correlation among endmembers and abundances. The model introduces endmember independence constraint for endmember estimation and exploits the low-rankness in abundance maps for abundance estimation. Experimental results show that the proposed algorithm is effective for hyperspectral unmixing.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Wanguang Yin, Youzhi Qu, Zhengming Ma, Quanying Liu
Summary: Tensor decomposition is an effective tool for learning multi-way structures and heterogeneous features from high-dimensional data. However, most methods fail to reveal the nonlinear structure within the data. To address this, we propose a novel method that uses hypergraphs to model complex connections among samples and employs factor matrices as low-dimensional representations.
Article
Geochemistry & Geophysics
Ziyang Guo, Anyou Min, Bing Yang, Junhong Chen, Hong Li, Junbin Gao
Summary: This article proposes a new sparse oblique-manifold (OB) NMF method inspired by matrix manifold theory, treating the abundance matrix as located on the oblique manifold to eliminate constraints and incorporate intrinsic Riemannian geometry. By using the Riemannian conjugated gradient (RCG) algorithm and multiplicative iterative rule, the proposed method not only improves solution accuracy but also achieves a faster convergence rate. Experimental results demonstrate the effectiveness and efficiency of the proposed method compared to state-of-the-art NMF methods in HU.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Computer Science, Information Systems
Anmin Fu, Zhenzhu Chen, Yi Mu, Willy Susilo, Yinxia Sun, Jie Wu
Summary: This paper presents a novel outsourced scheme for Non-negative Matrix Factorization (O-NMF) to alleviate clients' computing burden and address data privacy and verification issues when outsourcing NMF.
IEEE TRANSACTIONS ON SERVICES COMPUTING
(2022)
Article
Computer Science, Artificial Intelligence
Jiewen Guan, Bilian Chen, Xin Huang
Summary: Community detection aims to find densely connected communities in a network, which is a fundamental tool for various applications. Nonnegative matrix factorization (NMF)-based methods have gained attention, but they often neglect multihop connectivity patterns. In this article, we propose a novel method called multihop NMF (MHNMF) that considers these patterns and outperforms state-of-the-art methods.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Engineering, Electrical & Electronic
Bardia Yousefi, Clemente Ibarra Castanedo, Xavier P. V. Maldague
Summary: This study conducts a comparative analysis on low-rank matrix approximation methods in thermography and demonstrates the practicality and efficiency of semi-, convex-, and sparse-nonnegative matrix factorization methods for detecting subsurface thermal patterns. The experimental results show that these methods are effective in subsurface defect detection and distinguishing breast abnormalities in breast cancer screening data sets.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Automation & Control Systems
Yinan Li, Ruili Wang, Yuqiang Fang, Meng Sun, Zhangkai Luo
Summary: This article proposes a variable splitting based convolutive NMF algorithm to address the issues of low convergence rates, difficulty in reaching optimal solutions, and sparse results. Experimental results demonstrate the superiority of the proposed algorithm in terms of efficiency, optimal solutions, and sparsity.
IEEE TRANSACTIONS ON CYBERNETICS
(2022)
Article
Statistics & Probability
Xueying Liu, Jeremy Carter, Brad Ray, George Mohler
Summary: This study presents a spatial-temporal point process model for drug overdose clustering, integrating heterogeneous data sources to predict drug overdose death hotspots more accurately. The findings show significant excitation in drug and opioid overdose deaths, with a branching ratio ranging from 0.72 to 0.98.
ANNALS OF APPLIED STATISTICS
(2021)
