Search

Automatic data fusion is an important field of machine learning that has been increasingly studied. The objective is to improve the classification performance from several individual classifiers in terms of accuracy and stability of the results. This paper presents a comparative study on recent data fusion methods. The fusion step can be applied at early and/or late stages of the classification procedure. Early fusion consists of combining features from different sources or domains to form the observation vector before the training of the individual classifiers. On the contrary, late fusion consists of combining the results from the individual classifiers after the testing stage. Late fusion has two setups, combination of the posterior probabilities (scores), which is called soft fusion, and combination of the decisions, which is called hard fusion. A theoretical analysis of the conditions for applying the three kinds of fusion (early, late, and late hard) is introduced. Thus, we propose a comparative analysis with different schemes of fusion, including weaknesses and strengths of the state-of-the-art methods studied from the following perspectives: sensors, features, scores, and decisions. [ABSTRACT FROM AUTHOR]

It is demonstrated that data augmentation is a promising approach to reduce the size of the captured dataset required for training automatic road surface classifiers. The context is on-board systems for autonomous or semi-autonomous driving assistance: automatic power-assisted steering. Evidence is obtained by extensive experiments involving multiple captures from a 10-channel multisensor deployment: three channels from the accelerometer (acceleration in the X, Y, and Z axes); three microphone channels; two speed channels; and the torque and position of the handwheel. These captures were made under different settings: three worm-gear interface configurations; hands on or off the wheel; vehicle speed (constant speed of 10, 15, 20, 30 km/h, or accelerating from 0 to 30 km/h); and road surface (smooth flat asphalt, stripes, or cobblestones). It has been demonstrated in the experiments that data augmentation allows a reduction by an approximate factor of 1.5 in the size of the captured training dataset. [ABSTRACT FROM AUTHOR]

The use of surrogate data algorithms for blind detection of nonlinearities in multiple-echo ultrasonic signals is analyzed by means of simulations. Results show that surrogate data technique is a potentially powerful tool for blind detection of nonlinearities in multiple-echo ultrasonic signals.

Optimum detection is applied to ultrasonic signals corrupted with significant levels of grain noise. The aim is to enhance the echoes produced by the interface between the first and second layers of a dome to obtain interface traces in echo pulse B-scan mode. This is useful information for the restorer before restoration of the dome paintings. Three optimum detectors are considered: matched filter, signal gating, and prewhitened signal gating. Assumed models and practical limitations of the three optimum detectors are considered. The results obtained in the dome analysis show that prewhitened signal gating outperforms the other two optimum detectors.

Alpha integration methods have been used for integrating stochastic models and fusion in the context of detection (binary classification). Our work proposes separated score integration (SSI), a new method based on alpha integration to perform soft fusion of scores in multiclass classification problems, one of the most common problems in automatic classification. Theoretical derivation is presented to optimize the parameters of this method to achieve the least mean squared error (LMSE) or the minimum probability of error (MPE). The proposed alpha integration method was tested on several sets of simulated and real data. The first set of experiments used synthetic data to replicate a problem of automatic detection and classification of three types of ultrasonic pulses buried in noise (four-class classification). The second set of experiments analyzed two databases (one publicly available and one private) of real polysomnographic records from subjects with sleep disorders. These records were automatically staged in wake, rapid eye movement (REM) sleep, and non-REM sleep (three-class classification). Finally, the third set of experiments was performed on a publicly available database of single-channel real electroencephalographic data that included epileptic patients and healthy controls in five conditions (five-class classification). In all cases, alpha integration performed better than the considered single classifiers and classical fusion techniques. [ABSTRACT FROM AUTHOR]

In this paper we present a new method to obtain the density of solid composites using ultrasound. The method exploits the relation between density and acoustic impedance and only requires the measurement of a reference signal form the interface water-air. Once made in a controlled environment, it can be used for any material, and does not require any additional measurement or set-up, so that with a simple calculus it can be used with automatic ultrasonic measurements to calculate the density and therefore elastic constants of the composite for its complete characterization. The method is applied to 3 different materials: two composites based on epoxy and polyester resins respectively with unknown final properties and a common methacrylate sample used as reference. Results show good concordance between obtained and expected results. [ABSTRACT FROM AUTHOR]

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 to the detection context. Three optimization methods are presented: least mean square error, maximization of the area under the ROC curve, and minimization of the probability of error. Gradient algorithms are proposed for the three methods. Different experiments with simulated and real data are included. Simulated data consider the twodetector case to illustrate the factors influencing alpha integration and demonstrate the improvements obtained by score fusion with respect to individual detector performance. Two real data cases have been considered. In the first, multimodal biometric data have been processed. This case is representative of scenarios in which the probability of detection is to be maximized for a given probability of false alarm. The second case is the automatic analysis of electroencephalogram and electrocardiogram records with the aim of reproducing the medical expert detections of arousal during sleeping. This case is representative of scenarios in which probability of error is to be minimized. The general superior performance of alpha integration verifies the interest of optimizing the fusing parameters. [ABSTRACT FROM AUTHOR]

The detection and identification of internal defects in a material require the use of some technology that translates the hidden interior damages into observable signals with different signature-defect correspondences. We apply impact-echo techniques for this purpose. The materials are classified according to their defective status (homogeneous, one defect or multiple defects) and kind of defect (hole or crack, passing through or not). Every specimen is impacted by a hammer, and the spectrum of the propagated wave is recorded. This spectrum is the input data to a Bayesian classifier that is based on the modeling of the conditional probabilities with a mixture of Gaussians. The parameters of the Gaussian mixtures and the class probabilities are estimated using an extended expectation-maximization algorithm. The advantage of our proposal is that it is flexible, since it obtains good results for a wide range of models even under little supervision; e.g., it obtains a harmonic average of precision and recall value of 92.38% given only a 10% supervision ratio. We test the method with real specimens made of aluminum alloy. The results show that the algorithm works very well. This technique could be applied in many industrial problems, such as the optimization of the marble cutting process. [ABSTRACT FROM AUTHOR]

Missing traces in ground penetrating radar (GPR) B-scans (radargrams) may appear because of limited scanning resolution, failures during the acquisition process or the lack of accessibility to some areas under test. Four statistical interpolation methods for recovering these missing traces are compared in this paper: Kriging, Wiener structures, Splines and the expectation assuming an independent component analyzers mixture model (E-ICAMM). Kriging is an adaptation to the spatial context of the linear least mean squared error estimator. Wiener structures improve the linear estimator by including a nonlinear scalar function. Splines are a commonly used method to interpolate GPR traces. This consists of piecewise-defined polynomial curves that are smooth at the connections (or knots) between pieces. E-ICAMM is a new method proposed in this paper. E-ICAMM consists of computing the optimum nonlinear estimator (the conditional mean) assuming a non-Gaussian mixture model for the joint probability density in the observation space. The proposed methods were tested on a set of simulated data and a set of real data, and four performance indicators were computed. Real data were obtained by GPR inspection of two replicas of historical walls. Results show the superiority of E-ICAMM in comparison with the other three methods in the application of reconstructing incomplete B-scans. [ABSTRACT FROM AUTHOR]

Abstract: This paper introduces the application of independent component analysis mixture modelling (ICAMM) in non-destructive testing (NDT). The application consists of discriminating patterns for material quality control from homogeneous and defective materials inspected by impact-echo testing. This problem is modelled as a mixture of independent component analysis (ICA) models, representing a class of defective or homogeneous material by an ICA model whose parameters are learned from the impact-echo signal spectrum. These parameters define a kind of particular signature for the different defects. The proposed procedure is intended to exploit to the maximum the information obtained with the cost efficiency of only a single impact. To illustrate this capability, four levels of classification detail (material condition, kind of defect, defect orientation, and defect dimension) are defined, with the lowest level of detail having up to 12 classes. The results from several 3D finite element models and lab specimens of an aluminium alloy that contain defects of different shapes and sizes in different locations are included. The performance of the classification by ICA mixtures is compared with linear discriminant analysis (LDA) and with multi-layer perceptron (MLP) classification. We demonstrate that the mass spectra from impact-echo testing fit ICAMM, and we also show the feasibility of ICAMM to contribute in NDT applications. [Copyright &y& Elsevier]

Abstract: This paper presents a new procedure for learning mixtures of independent component analyzers. The procedure includes non-parametric estimation of the source densities, supervised–unsupervised learning of the model parameters, incorporation of any independent component analysis (ICA) algorithm into the learning of the ICA mixtures, and estimation of residual dependencies after training for correction of the posterior probability of every class to the testing observation vector. We demonstrate the performance of the procedure in the classification of ICA mixtures of two, three, and four classes of synthetic data, and in the classification of defective materials, consisting of 3D finite element models and lab specimens, in non-destructive testing using the impact-echo technique. The application of the proposed posterior probability correction demonstrates an improvement in the classification accuracy. Semi-supervised learning shows that unlabeled data can degrade the performance of the classifier when they do not fit the generative model. Comparative results of the proposed method and standard ICA algorithms for blind source separation in one and multiple ICA data mixtures show the suitability of the non-parametric ICA mixture-based method for data modeling. [Copyright &y& Elsevier]

The method uses ancillary data such as the pixel spectral similarity to enhance the quality of the fine-spatial resolution maps, thus improving the overall accuracy and prediction of the changed pixels. Innovative remote sensing image processing techniques have been progressively studied due to the increasing availability of remote sensing images, powerful techniques of data analysis, and computational power. Sub-pixel processing has emerged as an intense area of research considering the possibility of a pixel to belong to different classes in an image segmentation context. [Extracted from the article]

Many tasks in hyperspectral imaging, such as spectral unmixing and sub-pixel matching, require knowing how many substances or materials are present in the scene captured by a hyperspectral image. In this paper, we present an algorithm that estimates the number of materials in the scene using agglomerative clustering. The algorithm is based on the assumption that a valid clustering of the image has one cluster for each different material. After reducing the dimensionality of the hyperspectral image, the proposed method obtains an initial clustering using K-means. In this stage, cluster densities are estimated using Independent Component Analysis. Based on the K-means result, a model-based agglomerative clustering is performed, which provides a hierarchy of clusterings. Finally, a validation algorithm selects a clustering of the hierarchy; the number of clusters it contains is the estimated number of materials. Besides estimating the number of endmembers, the proposed method can approximately obtain the endmember (or spectrum) of each material by computing the centroid of its corresponding cluster. We have tested the proposed method using several hyperspectral images. The results show that the proposed method obtains approximately the number of materials that these images contain. [ABSTRACT FROM AUTHOR]

The essential step of surrogating algorithms is phase randomizing the Fourier transform while preserving the original spectrum amplitude before computing the inverse Fourier transform. In this paper, we propose a new method which considers the graph Fourier transform. In this manner, much more flexibility is gained to define properties of the original graph signal which are to be preserved in the surrogates. The complex case is considered to allow unconstrained phase randomization in the transformed domain, hence we define a Hermitian Laplacian matrix that models the graph topology, whose eigenvectors form the basis of a complex graph Fourier transform. We have shown that the Hermitian Laplacian matrix may have negative eigenvalues. We also show in the paper that preserving the graph spectrum amplitude implies several invariances that can be controlled by the selected Hermitian Laplacian matrix. The interest of surrogating graph signals has been illustrated in the context of scarcity of instances in classifier training. [ABSTRACT FROM AUTHOR]

Conventional partial correlation coefficients (PCC) were extended to the non-Gaussian case, in particular to independent component analysis (ICA) models of the observed multivariate samples. Thus, the usual methods that define the pairwise connections of a graph from the precision matrix were correspondingly extended. The basic concept involved replacing the implicit linear estimation of conventional PCC with a nonlinear estimation (conditional mean) assuming ICA. Thus, it is better eliminated the correlation between a given pair of nodes induced by the rest of nodes, and hence the specific connectivity weights can be better estimated. Some synthetic and real data examples illustrate the approach in a graph signal processing context. [ABSTRACT FROM AUTHOR]

Conventional partial correlation coefficients (PCC) were extended to the non-Gaussian case, in particular to independent component analysis (ICA) models of the observed multivariate samples. Thus, the usual methods that define the pairwise connections of a graph from the precision matrix were correspondingly extended. The basic concept involved replacing the implicit linear estimation of conventional PCC with a nonlinear estimation (conditional mean) assuming ICA. Thus, it is better eliminated the correlation between a given pair of nodes induced by the rest of nodes, and hence the specific connectivity weights can be better estimated. Some synthetic and real data examples illustrate the approach in a graph signal processing context.

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 linear restrictions are relaxed using several ICA mixture models (ICAMMs) obtaining a two-layer artificial neural network structure. This allows for dependence between sources of different classes, and thus, a myriad of multidimensional probability density functions can be accurate modeled. This paper proposes a new probabilistic distance (PDI) between the parameters learned for two ICAMMs. The PDI is computed explicitly, unlike the popular Kullback-Leibler divergence (KLD) and other similar metrics, removing the need for numerical integration. Furthermore, the PDI is symmetric and bounded within 0 and 1, which enables its use as a posterior probability in fusion approaches. In this paper, the PDI is employed for change detection by measuring the distance between two ICAMMs learned in consecutive time windows. The changes might be associated with relevant states from a process under analysis that are explicitly reflected in the learned ICAMM parameters. The proposed distance was tested in two challenging applications using simulated and real data: 1) detecting flaws in materials using ultrasounds and 2) detecting changes in electroencephalography signals from humans performing neuropsychological tests. The results demonstrate that the PDI outperforms the KLD in change-detection capabilities.