In this paper, we propose a novel diffusion robust variable step-size least mean square (DRVSS-LMS) algorithm that is insensitive to impulsive noise for distributed estimation in the network. Conventional diffusion least mean square algorithms are based on the assumption that the background noise obeys Gaussian distribution. However, the performances of these algorithms are severely degraded when impulsive noises occur in the network. By introducing the Huber objective function which can significantly suppress the effect of impulsive noise on estimation performances, we introduce a novel method to respectively deal with the abnormal nodes carrying data contaminated by impulsive noise and the normal nodes without being contaminated by impulsive noise. In addition, the proposed algorithm is assigned with variable step-sizes to further improve the performances of distributed estimation. Simulation results show that the proposed DRVSS-LMS algorithm can achieve both higher convergence rate and lower steady-state misadjustment than several existing robust diffusion LMS algorithms in the presence of impulsive noise.
5 January 2022 in IEEE Journal of Biomedical and Health Informatics
Objective: Preterm birth is the leading cause of neonatal morbidity and mortality. Early identification of high-risk patients followed by medical interventions is essential to the prevention of preterm birth. Based on the relationship between uterine contraction and the fundamental electrical activities of muscles, we extracted effective features from EHG signals recorded from pregnant women, and use them to train classifiers with the purpose of providing high precision in classifying term and preterm pregnancies. Methods: To characterize changes from irregularity to coherence of the uterine activity during the whole pregnancy, network representations of the original electrohysterogram (EHG) signals are established by applying the Horizontal Visibility Graph (HVG) algorithm, from which we extract network degree density and distribution, clustering coefficient and assortativity coefficient. Concerns on the interferences of different noise sources embedded in the EHG signal, we apply Short-Time Fourier Transform (STFT) to expand the original signal in the time-frequency domain. This allows a network representation and the extraction of related features on each frequency component. Feature selection algorithms are then used to filter out unrelated frequency components. We further apply the proposed feature extraction method to EHG signals available in the Term-Preterm EHG database (TPEHG), and use them to train classifiers. We adopt the Partition-Synthesis scheme which splits the original imbalanced dataset into two sets, and synthesizes artificial samples separately within each subset to solve the problem of dataset imbalance. Results: The optimally selected network-based features, not only contribute to the identification of the essential frequency components of uterine activities related to preterm birth, but also to improved performance in classifying term/preterm pregnancies, i.e., the SVM (Support Vector Machine) classifier trained with the available samples in the TPEHG gives sensitivity, specificity, overall accuracy, and $auc$ values as high as 0.89, 0.93, 0.91, and 0.97, respectively.
Blades icing will seriously affect the performance of wind turbines in respect to power loss and dynamic load increase. Blades icing detection technique becomes necessary to advance de-icing maintenance. Extracting effective features from SCADA data has become a challenging task during operating conditions under icing. Current research work lacks for the integration of physical information and insufficient analysis of icing dynamic evolution in the process of feature extraction. In order to eliminate these deficiencies, a multilevel feature extraction model combined with physical information and intelligent algorithm (MFT-PI) is proposed in this paper. First, features that characterize the severity of icing are extracted by power characteristic curve technology and feature selection technology in the transient feature extraction level. These techniques are driven by the information of power loss and icing mass growth. Second, a fully connected neural network integrated with triplet loss function is designed for extracting fusion feature, which considers the similarity of data distribution between early-icing stage and non-icing stage. Finally, the proposed feature extraction model is evaluated in two icing wind turbines. Results show that fusion features extracted from the proposed model is more stable, reliable and accurate by comparing the performance of other feature groups in different classifiers.
In this study, the authors propose a novel component ‐ wise variable step‐size (CVSS) diffusion distributed algorithm for estimating a specific parameter over sensor networks. The novelty of the CVSS algorithm is that step‐sizes vary from each other on different components at each iteration. They derive the steady‐state value of global mean‐square deviation (MSD) and relative MSD (RMSD). In the numerical simulations, they compare the proposed CVSS algorithm with several other least mean square (LMS) algorithms. Results show that, when compared with these other algorithms, the CVSS algorithm can effectively reduce steady‐state value and speed up convergence rate of RMSD while not sacrificing the convergence rate of MSD. Results also reveal that the proposed CVSS algorithm can achieve reduced difference of steady‐state values of relative estimation error on various components.
The quaternion Gaussian kernel is usually used when solving quaternion nonlinear problems. However, how to choose a proper value of kernel width is still an important issue. In most previous studies, the kernel width was set manually or estimated in advance by using Silvermans rule based on the sample distribution, which can easily degrade the performance of algorithms. In this brief, the variable kernel width quaternion kernel least mean squares algorithm (VKW-QKLMS) aims to develop an online technique for optimizing the kernel width of the quaternion kernel LMS (QKLMS) algorithm, in which the filter weight and the kernel width are alternately updated by using stochastic gradient algorithm. Simulation results show that the performance of the VKW-QKLMS algorithm does not depend on the selection of initial value of the kernel width. The VKW-QKLMS algorithm can usually achieve better performances than other competing algorithms. Only when the kernel width in the QKLMS algorithm is empirically selected as a proper value, the QKLMS algorithm can achieve much the same performance as that of the VKW-QKLMS algorithm.