A5018930444- Blind Source Separation Techniques
- Ultrasonics and Acoustic Wave Propagation
- Spectroscopy and Chemometric Analyses
- Geophysical Methods and Applications
- EEG and Brain-Computer Interfaces
- Speech and Audio Processing
- Neural Networks and Applications
- Underwater Acoustics Research
- Anomaly Detection Techniques and Applications
- Non-Destructive Testing Techniques
- Fault Detection and Control Systems
- Advanced Chemical Sensor Technologies
- Structural Health Monitoring Techniques
- Imbalanced Data Classification Techniques
- Flow Measurement and Analysis
- Music and Audio Processing
- Fire Detection and Safety Systems
- Advanced Statistical Process Monitoring
- Infrastructure Maintenance and Monitoring
- Advanced Adaptive Filtering Techniques
- Remote Sensing in Agriculture
- Remote-Sensing Image Classification
- Advanced Statistical Methods and Models
- Target Tracking and Data Fusion in Sensor Networks
- Fire effects on ecosystems
Universitat Politècnica de València
2015-2024
Vera C. Rubin Observatory
2024
University of Oklahoma
2021
Multimedia University
2010-2013
Universitat de València
2007-2008
Universidad Politécnica de Madrid
1988-2005
University of the Andes
2005
European Telecommunications Standards Institute
1985-2004
University of Missouri–St. Louis
1971
In this article we carry out a comparison between early (feature) and late (score) multimodal fusion, for the two-class problem. The is made first from general perspective, then specific mathematical analysis. Thus, deduce error probability expressions uncorrelated correlated multivariate Gaussian distribution, assuming perfect model knowledge (Bayes rates). We also corresponding when to be learned finite training set, demonstrating its convergence Bayes rates as set size goes infinite....
Optimum detection is applied to ultrasonic signals corrupted with significant levels of grain noise. The aim enhance the echoes produced by interface between first and second layers a dome obtain traces in echo pulse B-scan mode. This useful information for restorer before restoration paintings. Three optimum detectors are considered: matched filter, signal gating, prewhitened gating. Assumed models practical limitations three considered. results obtained analysis show that gating...
We present a new method for fusing scores corresponding to different detectors (two-hypotheses case). It is based on alpha integration, which we have adapted the detection context. Three optimization methods are presented: least mean square error, maximization of area under ROC curve, and minimization probability error. Gradient algorithms proposed three methods. Different experiments with simulated real data included. Simulated consider two-detector case illustrate factors influencing...
In this paper, a theoretical learning curve is derived for the multi-class Bayes classifier. This fits general multivariate parametric models of class-conditional probability density. The derivation uses proxy approach based on analyzing convergence statistic which proportional to posterior true class. By doing so, depends only training set size and dimension feature vector; it does not depend model parameters. Essentially, provides an estimate reduction in excess error that can be obtained...
This paper presents a theoretical derivation of two new graph-based regularization methods for fusing the individual results multiple detectors (two-class classifiers). The proposed approach considers linear combination detector statistics and its extension to general nonlinear fusion method known as α-integration. A cost function that includes mean-square error term is minimized. inclusion term, which based on graph signal processing, reduces dispersion fused statistics, thus improves...
Success in supervised learning is constrained by availability of an adequate labeled data sample for training. The problem a complete labeling every the training dataset can be alleviated allowing semi-complete way so called semi-supervised learning. In this paper, we investigate performance imbalanced classification problems. Augmentation class limited applied lowering variance estimate using subrogation method. We analyze effect augmentation several simulated and experimental scenarios...
Alpha integration methods have been used for integrating stochastic models and fusion in the context of detection (binary classification). Our work proposes separated score (SSI), a new method based on alpha to perform soft scores multiclass classification problems, one most common problems automatic classification. Theoretical derivation is presented optimize parameters this achieve least mean squared error (LMSE) or minimum probability (MPE). The proposed was tested several sets simulated...
The majority of experiments in fundamental science today are designed to be multi-purpose: their aim is not simply measure a single physical quantity or process, but rather enable increased precision the measurement number different observable quantities natural system, extend search for new phenomena, exclude larger phase space candidate theories. Most time, combination above goals pursued; this breadth scope adds layer complexity already demanding task designing apparatus an optimal way,...
Banks collect large amount of historical records corresponding to millions credit cards operations, but, unfortunately, only a small portion, if any, is open access. This because, e.g., the include confidential customer data and banks are afraid public quantitative evidence existing fraud operations. paper tackles this problem with application surrogate techniques generate new synthetic card data. The quality multivariate guaranteed by constraining them have same covariance, marginal...
Many tasks in hyperspectral imaging, such as spectral unmixing and sub-pixel matching, require knowing how many substances or materials are present the scene captured by a image. In this paper, we an algorithm that estimates number of using agglomerative clustering. The is based on assumption valid clustering image has one cluster for each different material. After reducing dimensionality image, proposed method obtains initial K-means. stage, densities estimated Independent Component...
Automatic data fusion is an important field of machine learning that has been increasingly studied. The objective to improve the classification performance from several individual classifiers in terms accuracy and stability results. This paper presents a comparative study on recent methods. step can be applied at early and/or late stages procedure. Early consists combining features different sources or domains form observation vector before training classifiers. On contrary, results after...
Fraud detection is a critical problem affecting large financial companies that has increased due to the growth in credit card transactions. This paper presents new method for automatic of frauds transactions based on non-linear signal processing. The proposed consists following stages: feature extraction, training and classification, decision fusion, result presentation. Discriminant-based classifiers an advanced non-Gaussian mixture classification are employed distinguish between legitimate...
Missing traces in ground penetrating radar (GPR) B-scans (radargrams) may appear because of limited scanning resolution, failures during the acquisition process or lack accessibility to some areas under test. Four statistical interpolation methods for recovering these missing are compared this paper: Kriging, Wiener structures, Splines and expectation assuming an independent component analyzers mixture model (E-ICAMM). Kriging is adaptation spatial context linear least mean squared error...
The detection and identification of internal defects in a material require the use some technology that translates hidden interior damages into observable signals with different signature-defect correspondences. We apply impact-echo techniques for this purpose. materials are classified according to their defective status (homogeneous, one defect or multiple defects) kind (hole crack, passing through not). Every specimen is impacted by hammer, spectrum propagated wave recorded. This input...
Independent component analysis (ICA) is a blind source separation technique where data are modeled as linear combinations of several independent non-Gaussian sources. The independence and restrictions relaxed using ICA mixture models (ICAMMs) obtaining two-layer artificial neural network structure. This allows for dependence between sources different classes, thus, myriad multidimensional probability density functions can be accurate modeled. paper proposes new probabilistic distance (PDI)...