Jorge Gosálbez

Optimum detection of ultrasonic echoes applied to the analysis of the first layer of a restored dome

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.

Measurement of cement porosity by centroid frequency profiles of ultrasonic grain noise

In this paper, we propose a technique for material characterization by using centroid frequency profiles (CFP) of ultrasound echo signals. These echo signals are composed by grain noise due to the superposition of many small echoes from the inner microstructure plus observation noise. A CFP indicates the centroid frequency dependence on depth, corresponding to power spectrum density assessments at different depths. We show in the paper the relation between the mean and variance of the CFP and the grain-to-observation-noise-ratio (GOR) at every depth. The GOR depends on the material ultrasound attenuation, so that CFP may be used for material characterization. Although we consider here the estimation of cement paste porosity, the proposed technique may have general applicability. Cement paste is the main component of mortar and concrete. Therefore, cement porosity is an important problem because the vulnerability (and thence the durability) of these construction materials to external agents depends heavily on it. Experiments have been made to show the correlation between cement paste porosity and a penetration parameter obtained from the CFP.

Material grain noise analysis by using higher-order statistics

We have derived equations relating to the cumulants of the backscattered signal to material and transducer parameters. Then, we proposed a practical method to estimate the material grain moments from estimates of the cumulants at some particular values. The proposed model and techniques are verified on some phantoms having different scatterer density and grain sizes.

Fast communication: Detection of signals of unknown duration by multiple energy detectors

An extension of the classical energy detector is proposed to deal with the case of unknown signal duration. Multiple energy detectors are applied to partitions of the original observation interval; presence of signal is decided if at least one of the detectors is in favor of it. We have derived the corresponding probabilities of false alarm and detection for a particular strategy of successive segmentations of the original interval, thus obtaining a layered structure of energy detectors. One key problem is that individual decisions obtained from the multiple energy detectors are statistically dependent, thus complicating the derivation of the overall probabilities of detection and false alarm. ROC curves have been computed, showing significant improvements in detectability when there is a large mismatch between the duration of the observation interval and the actual duration of the signal. This can be especially interesting in the framework of novelty detection where specific parameters of the signal, like duration, are totally unknown.

An extended energy detector for non-Gaussian and non-independent noise

Energy detectors are optimum to detect uncorrelated Gaussian signals or generalized likelihood ratio tests to detect completely unknown signals; in both cases, background noise must be uncorrelated Gaussian. However, energy detectors degrade when background noise is non-independent and non-Gaussian. An extension is presented in this paper to deal with this situation. Independence is achieved by means of a matrix linear transformation derived from independent component analysis. Non-Gaussianity is avoided by applying a scalar non-linear function to every element of the linearly transformed observation vector. Practical procedures for estimating the linear and nonlinear transformations are given in the paper. A SNR enhancement factor has been defined for the weak signal case, which is indicative of the expected improvement of the proposed extension. Some simulations illustrate the achieved improvements.

ICA Model Applied to Multichannel Non-destructive Evaluation by Impact-Echo

This article presents an ICA model for applying in Non Destructive Testing by Impact-Echo. The approach consists in considering flaws inside the material as sources for blind separation using ICA. A material is excited by a hammer impact and a convolutive mixture is sensed by a multichannel system. Obtained information is used for classifying in defective or non defective material. Results based on simulation by finite element method are presented, including different defect geometry and location.

Acoustic detection and classification using temporal and frequency multiple energy detector features

The problem of acoustic detection and recognition is of particular interest in surveillance applications, especially in noisy environments with sound sources of different nature. Therefore, we present a multiple energy detector (MED) structure which is used to extract a new set of features for classification, called frequency MED (FMED) and combined MED (CMED). The focus of this paper is to compare these two novel feature sets with the commonly used MFCC and to evaluate their performance in a general sound classification task with different acoustic sources and adverse noise conditions. The promising results obtained show that, in low SNR, the proposed CMED features work significantly better than the MFCC.

Improving Detection of Acoustic Signals by Means of a Time and Frequency Multiple Energy Detector

Standard energy detectors (ED) are optimum to detect unknown signals in presence of uncorrelated Gaussian noise. However, in real applications the signal duration and bandwidth are unpredictable and this fact can considerably degrade the detection performance if the appropriate observation vector length is not correctly selected. Therefore, a multiple energy detector (MED) structure is applied in the time as well as in the frequency domain and it is evaluated in real acoustic scenarios. The results obtained demonstrate the robustness of the MED structure and a performance improvement in comparison to the standard ED.

Neural networks for defect detection in non-destructive evaluation by sonic signals

This paper presents an application of neural networks in pattern recognition of defects in sonic signals from non-destructive evaluation by multichannel impact-echo. The problem approached consists in allocating parallelepiped-shape materials in four levels of classifications defining material condition (homogeneous or defective), kind of defects (holes and cracks), defect orientation, and defect dimension. Various signal features as centroid frequency, attenuation and amplitude of the principal frequency are estimated per channel and processed by PCA and feature selection methods to reduce dimensionality. Results for simulations and experiments applying Radial Basis Function, Multilayer Perceptron and Linear Vector Quantization neural networks are presented. Neural networks obtain good performance in classifying several 3D finite element models and specimens of aluminum alloy.

Optimized ultrasonic attenuation measures for non-homogeneous materials.

In this paper the study of frequency-dependent ultrasonic attenuation in strongly heterogeneous materials is addressed. To determine the attenuation accurately over a wide frequency range, it is necessary to have suitable excitation techniques. Three kinds of transmitted signals have been analysed, grouped according to their bandwidth: narrowband and broadband signals. The mathematical formulation has revealed the relation between the distribution of energy in their spectra and their immunity to noise. Sinusoidal and burst signals have higher signal-to-noise ratios (SNRs) but need many measurements to cover their frequency range. However, linear swept-frequency signals (chirp) improve the effective bandwidth covering a wide frequency range with a single measurement and equivalent accuracy, at the expense of a lower SNR. In the case of highly attenuating materials, it is proposed to use different configurations of chirp signals, enabling injecting more energy, and therefore, improving the sensitivity of the technique without a high time cost. Thus, if the attenuation of the material and the sensitivity of the measuring equipment allows the use of broadband signals, the combination of this kind of signal and suitable signal processing results in an optimal estimate of frequency-dependent attenuation with a minimum measurement time.