Search

In this work we introduce a new approach for robust image segmentation. The idea is to combine two strategies within a Bayesian framework. The first one is to use a Markov Random Field (MRF), which allows to introduce prior information with the purpose of preserve the edges in the image. The second strategy comes from the fact that the probability density function (pdf) of the likelihood function is non Gaussian or unknown, so it should be approximated by an estimated version, and for this, it is used the classical non-parametric or kernel density estimation. This two strategies together lead us to the definition of a new maximum a posteriori (MAP) approach based on the minimization of the entropy of the estimated pdf of the likelihood function and the MRF at the same time, named MAP entropy estimator (MAPEE). Some experiments were conducted for different kind of images degraded with impulsive noise and other non-Gaussian distributions, where the segmentation results are very satisfactory comparing them with respect to recent robust approaches based on the fuzzy c-means (FCM) segmentation.

Training procedures of a deep neural network are still an area with ample research possibilities and constant improvement either to increase its efficiency or its time performance. One of the lesser-addressed components is its objective function, which is an underlying aspect to consider when there is the necessity to achieve better error rates in the area of automatic speech recognition. The aim of this paper is to present two new variations of the frame-level cost function for training a deep neural network with the purpose of obtaining superior word error rates in speech recognition applied to a case study in Spanish. The first proposed function is a fusion between the boosted cross-entropy and the so called cross-entropy/log-posterior-ratio. The main idea is to jointly emphasize the prediction of difficult/crucial frames provided by a boosting factor and at the same time enlarge the distance between the target senone and its closest competitor. The second proposal is a fusion between the non-uniform mapped cross-entropy and the cross-entropy/log-posterior-ratio. This function utilizes both the mapped function to enhance the frames that have ambiguity in their belonging to specific senones and the log-posterior-ratio with the purpose of separating the target senone against the most competing tied tri-phone state. The proposed approaches are compared against those frame-level cost functions discussed in the state of the art. This comparative has been made by using a personalized mid-vocabulary speaker-independent voice corpus. This dataset is employed for the recognition of digit strings and personal name lists in Spanish from the northern central part of Mexico on a connected-words phone dialing task. A relative word error rate improvement of 15.14% and 12.30% is obtained with the two proposed approaches, respectively, against the plain well-established cross-entropy loss function.

Amino acid complexes of zirconia represent an entirely new class of materials that were synthesized and studied for the first time for the decontamination of fluoride ion containing aqueous solutions. Glutamic and aspartic acid complexes of zirconia assembled with thin carbon (stacked graphene oxide) platelets deriving from graphite oxide (GO) were synthesized by a two-step method to prepare adsorbents. The characterization of the complexes was carried out using infrared spectroscopy to determine the functional groups and the types of interaction between the composites and fluoride ions. To reveal the mechanisms and extent of adsorption, two types of batch adsorption measurements were performed: (i) varying equilibrium fluoride ion concentrations to construct adsorption isotherms at pH = 7 in the absence of added electrolytes and (ii) using fixed initial fluoride ion concentrations (10 mg/L) with a variation of either the pH or the concentration of a series of salts that potentially interfere with adsorption. The experimental adsorption isotherms were fitted by three different theoretical isotherm equations, and they are described most appropriately by the two-site Langmuir model for both adsorbents. The adsorption capacities of Zr-glutamic acid-graphite oxide and Zr-aspartic acid-graphite oxide are 105.3 and 101.0 mg/g, respectively. We found that two distinct binding modes are combined in the Zr-amino acid complexes: at low solution concentrations, F− ions are preferentially adsorbed by coordinating to the surface Zr species up to a capacity of ca. 10 mg/g. At higher concentrations, however, large amounts of fluoride ions may undergo anion exchange processes and physisorption may occur on the positively charged ammonium moieties of the interfacially bound amino acid molecules. The high adsorption capacity and affinity of the studied dicarboxylate-type amino acids demonstrate that amino acid complexes of zirconia are highly variable materials for the safe and efficient capture of strong Lewis base-type ions such as fluoride.

The correct phase unwrapping in the presence of noise is a difficult problem and it has been of great interest since the last decades as observed in the literature. This is due to the spatial dependency that exists in traditional 2-D phase-unwrapping algorithms and, as a consequence, spreads errors in the reconstructed phase map. To solve this, algorithms known as temporal phase unwrapping (TPU) have been proposed; here the intensity of a fringe image sequence changes as a function of time, and therefore the elements of the phase-map are independent of each other. Thus, the 3-D measurement vision system by projection of fringe images presented here consists in capturing a sequence of images with sinusoidal fringes deformed by the height of the object. Then, the phase is processed by using a TPU technique based the one-dimensional Continuous Wavelet Transform (CWT). It should be noted that the use of CWT in one dimension is to analyze the intensity variations of the temporal sampled phase images, and the importance of this is in the detection of the frequency from which the phase is obtained as a linear function by means of an unwrapping in one dimension. We present simulated experiments and some real applications in digital archaeology and zoological morphometrics that validate the proposal.

A precise measurement in plants is necessary to know in detail the functions and processes of these living beings, which are a fundamental part for the development of life on the planet. The optimal growth conditions for any plant generated in a controlled atmosphere are very helpful to make measurements of the variable that may be the highest hierarchy: light. This paper describes the design of a growth sphere for optical measurements in plants (GSOMP), which fulfills two primary functions; generate the optimal conditions for the growth of the plant during its whole life cycle and perform incident light absorption measurements taking advantage of the geometric shape (spherical) and the coating of the device surface, so it is possible to obtain the light absorption spectrum in each phenological stage of any plant. Due to the aforementioned, this device is unique and represents a very useful option for research and experimentation in vegetables.

Temporal techniques for fringe projection profilometry, in particular the multi-frequency class, require the projection of a sequence of fringes with preset frequencies. In this paper, we analyze the generation of fringes that offers better conditions to process them through the modification of the fringe-frequency range, which may reduce measurement errors. This will also serve to expand the possibilities of being able to process the fringe information through different signal-processing methods. Additionally, in this work, we derive a first-approximation frequency-to-height conversion formula and describe an algorithm to process the fringe image sequence through the Continuous Wavelet-Transform (CWT). Experimental results with simulations and real applications are presented as well.

The aim of this paper is to illustrate an overview of the automatic speech recognition (ASR) module in a spoken dialog system and how it has evolved from the conventional GMM-HMM (Gaussian mixture model - hidden Markov model) architecture toward the recent nonlinear DNN-HMM (deep neural network) scheme. GMMs have dominated for a long time the baseline of speech recognition, but in the past years with the resurgence of artificial neural networks (ANNs), the former models have been surpassed in most recognition tasks. An outstanding consideration for ANNs-based acoustic model is the fact that their weights can be adjusted in two training steps: i) initialization of the weights (with or without pre-training) and ii) fine-tuning. To exemplify these frameworks, a case study is realized by using the Kaldi toolkit, employing a mid-vocabulary with a personalized speaker-independent voice corpus on a connected-words phone dialing environment operated for recognition of digit strings and personal name lists in Spanish from Mexico. The obtained results show a reasonable accuracy in the speech recognition performance through the DNN acoustic modeling. A word error rate (WER) of 1.49% for context-dependent DNN-HMM is achieved, providing a 30% relative improvement with regard to the best GMM-HMM result in these experiments (2.12% WER).

The present work proposes a review and comparison of different Kernel functionals and neighborhood geometry for Nonlocal Means (NLM) in the task of digital image filtering. Some different alternatives to change the classical exponential kernel function used in NLM methods are explored. Moreover, some approaches that change the geometry of the neighborhood and use dimensionality reduction of the neighborhood or patches onto principal component analysis (PCA) are also analyzed, and their performance is compared with respect to the classic NLM method. Mainly, six approaches were compared using quantitative and qualitative evaluations, to do this an homogeneous framework has been established using the same simulation platform, the same computer, and same conditions for the initializing parameters. According to the obtained comparison, one can say that the NLM filtering could be improved when changing the kernel, particularly for the case of the Tukey kernel. On the other hand, the excellent performance given by recent hybrid approaches such as NLM SAP, NLM PCA (PH), and the BM3D SAPCA lead to establish that significantly improvements to the classic NLM could be obtained. Particularly, the BM3D SAPCA approach gives the best denoising results, however, the computation times were the longest.

A new technique based on the Band-Limited Phase-Only Correlation (BLPOC) function to deal with acoustic individual identification is proposed in this paper. This is a biometric technique suitable for limited data individual bird identification. The main advantage of this new technique, in contrast to traditional algorithms where the use of large-scale datasets is assumed, is its ability to identify individuals by the use of only two samples from the bird species. The proposed technique has two variants (depending on the method used to analyze and extract the bird vocalization from records): automatic individual verification algorithm and semi-automatic individual verification algorithm. The evaluation of the automatic algorithm shows an average precision that is over 80% for the identification comparatives. It is shown that the efficiencies of the algorithms depend on the complexity of the vocalizations.

In this paper, it is proposed a gaussian convolution-based fringe pattern denoising method. As will be shown, this method is robust enough compared with some of the most outstanding methods in the literature. Additionally, the proposed method overcomes the problem of underperformance in low-frequency fringes that is common in most oriented filtering methods, while keeping the great advantages of convolution-based filters. The advantages of the proposed denoising method will be demonstrated with experiments realized over synthetic and real fringe patterns, and comparing the performance with four representative methods, already reported.

In this paper, we present an alternative technique for the demodulation of single interferograms using a sliding 2-D continuous wavelet transform (2-D CWT) method. The sliding strategy proposed in this work is used with two purposes: to reduce the processing time when the 2-D CWT is applied, and to solve the problem of the phase ambiguity when closed fringes are present. Experimental results with real and simulated interferograms show that the proposed multiresolution method is a proper alternative for many applications of interferogram demodulation with closed fringes.

The aim of this paper is to present two new variations of the frame-level cost function for training a deep neural network in order to achieve better word error rates in speech recognition. Minimization functions of a neural network are salient aspects to deal with when researchers are working on machine learning, and hence their improvement is a process of constant evolution. In the first proposed method, the conventional cross-entropy function can be mapped to a nonuniform loss function based on its corresponding extropy (a complementary dual function), enhancing the frames that have ambiguity in their belonging to specific senones (tied-triphone states in a hidden Markov model). The second proposition is a fusion of the proposed mapped cross-entropy and the boosted cross-entropy function, which emphasizes those frames with low target posterior probability. The developed approaches have been performed by using a personalized mid-vocabulary speaker-independent voice corpus. This dataset is employed for recognition of digit strings and personal name lists in Spanish from the northern central part of Mexico on a connected-words phone dialing task. A relative word error rate improvement of 12.3% and 10.7% is obtained with the two proposed approaches, respectively, regarding the conventional well-established crossentropy objective function.

The aim of this paper is to exhibit a comparative case study of the conventional speech recognition GMM-HMM (Gaussian mixture model — hidden Markov model) architecture and the recent model based on deep neural networks. During years the GMM approach has controlled the speech recognition tasks, however it has been surpassed with the resurgence of artificial neural networks. To exemplify these acoustic modeling frameworks, a case study has been conducted by using the Kaldi toolkit, employing a personalized speaker-independent mid-vocabulary voice corpus for recognition of digit strings and personal name lists in latin spanish on a connected-words phone dialing task. The speech recognition accuracy obtained in the results shows a better word error rate by using the DNN acoustic modeling. A 20.71% relative improvement is obtained with DNN-HMM models (3.33% WER) in respect to the lowest GMM-HMM rate (4.20% WER).

The present work proposes a review and comparison of different Nonlocal Means (NLM) methods in the task of digital image filtering. Some different alternatives to change the classical exponential kernel function used in NLM methods are explored. Moreover, some approaches that change the geometry of the neighborhood and use dimensionality reduction of the neighborhood or patches onto principal component analysis (PCA) are also analyzed, and their performance is compared with respect to the classic NLM method. Mainly, six approaches were compared using quantitative and qualitative evaluations, to do this an homogenous framework has been established using the same simulation platform, the same computer, and same conditions for the initializing parameters. One will notice that particularly, the BM3D SAPCA approach gives the best denoising results, but in contrast, the computation times of this method were the longest.

Aims: To determine the metallicities of 113 Southern Hemisphere Vega-like candidate stars in relation to the Exoplanet host group and field stars. Methods: We applied two spectroscopic methods of abundance determinations: equivalent width measurements together with the ATLAS9 (Kurucz 1993) model atmospheres and the WIDTH9 program, and a comparison of observed spectra with the grid of synthetic spectra of Munari et al. (2005). Results: For the Vega-like group, the metallicities are indistinguishable from those of field stars not known to be associated with planets or disks. This result is quite different from the metallicities of Exoplanet host stars which are metal-rich in comparison to field stars., 25 pages, 4 figures

This paper describes an alternative methodology to determine the thermal effusivity of a liquid sample using the recently proposed electropyroelectric technique, without fitting the experimental data with a theoretical model and without having to know the pyroelectric sensor related parameters, as in most previous reported approaches. The method is not absolute, because a reference liquid with known thermal properties is needed. Experiments have been performed that demonstrate the high reliability and accuracy of the method with measurement uncertainties smaller than 3%.

The present work introduces a new alternative to change the classical exponential kernel function used in Nonlocal Means (NLM) methods to deal with digital image filtering. The idea is based on the premise that making a good selection or estimation of the bandwidth parameter h is difficult and there are some other kernels which have another equivalent parameters to be selected into a more easiest way. A First method is obtained, when using an optimal manner proposed in nonparametric estimation literature to estimate h to tune the exponential kernel function. And a second proposed method, is to change the exponential function by the so called Hilbert function where one must to choose a parameter d. This Hilbert function is used for the first time in the NLM framework. Finally, the obtained filtering results reveals, that the NLM Hilbert kernel approach gives similar performance to other approaches according to recent reported results in literature, and the first proposed methodology is restricted by the estimation of h.

Friction in mechanisms is a phenomenon that degrades the performance of control systems. Among the various ingredients present in the friction phenomenon, the viscous effect is particularly harmful. In order to deal with viscous friction, which has been recognized during experimental tests to be a complex function of the operating temperature, this paper proposes a smooth adaptive viscous friction compensator using a non-linear observer based on the Dahl model of friction. The performance of the proposed control scheme is illustrated through experiments on a direct-drive servo motor.

Four diuretic drugs banned in sport (amiloride, triamterene, bendroflumethiazide and bumetanide) have been separated by capillary zone electrophoresis (CZE) and detected using conventional fluorescence spectrometry. The effect of pH on electrophoretic parameters such as migration time, peak efficiency and peak height is discussed. Complete separation of the four drugs is achieved in less than 8 min at pH 8. No interference due to endogenous urine components is observed and thus direct urine analysis is feasible. Analytical figures of merit including precision and limits of detection are presented. Limits of detection range between 0.5 fmol for triamterene and 21.6 fmol for bumetanide.

Reliable and accurate testing methods are essential to guiding the polishing process during the figuring of optical telescope mirrors. With the natural advancement of technology, the procedures and instruments used to carry out this delicate task have consistently increased in sensitivity, but also in complexity and cost. Fortunately, throughout history, the Foucault knife-edge test has shown the potential to measure transverse aberrations in the order of the wavelength, mainly when described in terms of physical theory, which allows a quantitative interpretation of its characteristic shadowmaps. Our previous publication on this topic derived a closed mathematical formulation that directly relates the knife-edge position with the observed irradiance pattern. The present work addresses the quite unexplored problem of the wavefront's gradient estimation from experimental captures of the test, which is achieved by means of an optimization algorithm featuring a proposed ad hoc cost function. The partial derivatives thereby calculated are then integrated by means of a Fourier-based algorithm to retrieve the mirror's actual surface profile. To date and to the best of our knowledge, this is the very first time that a complete mathematical-grounded treatment of this optical phenomenon is presented, complemented by an image-processing algorithm which allows a quantitative calculation of the corresponding slope at any given point of the mirror's surface, so that it becomes possible to accurately estimate the aberrations present in the analyzed concave device just through its associated foucaultgrams.

From the Bayesian regularization theory we derive a quadratic cost-function for integrating wave-front gradient fields. In the proposed cost-function, the term of conditional distribution uses a central-differences model to make the estimated function well consistent with the observed gradient field. As will be shown, the results obtained with the central-differences model are superior to the results obtained with the backward-differences model, commonly used in other integration techniques. As a regularization term we use an isotropic first-order differences Markov Random-Field model, which acts as a low-pass filter reducing the errors caused by the noise. We present simulated and real experiments of the proposal applied in the Foucault test, obtaining good results.

In this work, a novel model of Markov Random Field (MRF) is introduced. Such a model is based on a proposed Semi-Huber potential function and it is applied successfully to image segmentation in presence of noise. The main difference with respect to other half-quadratic models that have been taken as a reference is, that the number of parameters to be tuned in the proposed model is smaller and simpler. The idea is then, to choose adequate parameter values heuristically for a good segmentation of the image. In that sense, some experimental results show that the proposed model allows an easier parameter adjustment with reasonable computation times.

The present chapter illustrates the use of some recent alternative methods to deal with digital image filtering and restoration. This collection of methods is inspired on the use of Markov Random Fields (MRF), which introduces prior knowledge of information that will allow, more efficiently, modeling the image acquisition process. The methods based on the MRF are analyzed and proposed into a Bayesian framework and their principal objective is to eliminate noise and some effects caused by excessive smoothness on the reconstruction process of images which are rich in contours or edges. In order to preserve object edges into the image, the use of certain convexity criteria into the MRF is proposed obtaining adequate weighting of cost functions in cases where discontinuities are remarked and, even better, for cases where such discontinuities are very smooth.