Miguel Mora-González

Typical testors generation based on an evolutionary algorithm

Typical testors are useful for both feature selection and feature relevance determination in supervised classification problems. However, reported algorithms that address the problem of finding the set of all typical testors have exponential complexity. In this paper, we propose to adapt an evolutionary method, the Hill-Climbing algorithm, with an acceleration operator in mutation process, to address this problem in polinomial time. Experimental results with the method proposed are presented and compared, in efficiency, with other methods, namely, Genetic Algorithms (GA) and Univariate Marginal Distribution Algorithm (UMDA).

A fast implementation of the CT_EXT algorithm for the testor property identification

Typical testors are a useful tool for both feature selection and for determining feature relevance in supervised classication problems. Nowadays, generating all typical testors of a training matrix is computationally expensive; all reported algorithms have exponential complexity, depending mainly on the number of columns in the training matrix. For this reason, different approaches such as sequential and parallel algorithms, genetic algorithms and hardware implementations techniques have been developed. In this paper, we introduce a fast implementation of the algorithm CT_EXT (which is one of the fastest algorithms reported) based on an accumulative binary tuple, developed for generating all typical testors of a training matrix. The accumulative binary tuple implemented in the CT_EXT algorithm, is a useful way to simplifies the search of feature combinations which fulfill the testor property, because its implementation decreases the number of operations involved in the process of generating all typical testors. In addition, experimental results using the proposed fast implementation of the CT_EXT algorithm and the comparison with other state of the art algorithms that generated typical testors are presented.

An evolutionary algorithm with acceleration operator to generate a subset of typical testors

The proposed Hill-Climbing algorithm incorporates an acceleration operator.The acceleration operator improves the exploration capability of the algorithm.Local search is more adequate than global search for generating typical testors.The proposed algorithm has better performance than other heuristics reported. This paper is focused on introducing a Hill-Climbing algorithm as a way to solve the problem of generating typical testors - or non-reducible descriptors - from a training matrix. All the algorithms reported in the state-of-the-art have exponential complexity. However, there are problems for which there is no need to generate the whole set of typical testors, but it suffices to find only a subset of them. For this reason, we introduce a Hill-Climbing algorithm that incorporates an acceleration operation at the mutation step, providing a more efficient exploration of the search space. The experiments have shown that, under the same circumstances, the proposed algorithm performs better than other related algorithms reported so far.

Phase recovery from a single interferogram with closed fringes by phase unwrapping

We describe a new algorithm for phase determination from a single interferogram with closed fringes based on an unwrapping procedure. Here we use bandpass filtering in the Fourier domain, obtaining two wrapped phases with sign changes corresponding to the orientation of the applied filters. An unwrapping scheme that corrects the sign ambiguities by comparing the local derivatives is then proposed. This can be done, assuming that the phase derivatives do not change abruptly among adjacent areas as occurs with smooth continuous phase maps. The proposed algorithm works fast and is robust against noise, as demonstrated in experimental and simulated data.

Radon transform algorithm for fingerprint core point detection

This article presents an innovative technique for solving the problem of finding the core within a fingerprint. The Radon transform and a tree clustering algorithm were key to locating the coordinates of the core. Binarization and high-pass filtering processes to improve the contrast in fingerprints are proposed. The core of a fingerprint is located in the geometric cross section of maxima and minima in the Radon transforms at 0° and 90°. The technique is very stable, since it only presents difficulties when the fingerprint core is located on the edges of the image or is nonexistent.

Improving Sensitivity and Specificity in Breast Cancer Detection Using Raman Spectroscopy and Bayesian Classification

Automated real-time tissue assessment using Raman spectroscopy for breast cancer detection is feasible; however, long-term specificity and sensitivity as reported so far in the literature can still be improved by a more reliable algorithm for breast cancer detection. Applying automatic reduction of background fluorescence to the Raman spectra and Bayesian classification on 18 discriminant Raman bands in the range of 1200–1800 cm−1, we achieved 100% accuracy in classifying breast biopsies of healthy and cancerous tissues, making it suitable for automated breast cancer diagnosis and appropriate for long-term use in real time in a surgery room or research scenarios. The long-term reliability of this approach was cross-validated using three methods: resubstitution, leave-one-out, and holdout. The holdout method has the potential of estimating the upper bound of classification error probability of the Bayesian classifier; the holdout method allowed us to perform 50,000 classification trials with only three misclassifications, which demonstrates that the high performance in sensitivity and specificity will be retained in future applications of this approach for breast cancer detection.

The Ronchi test using a liquid crystal display as a phase grating

The Ronchi test with a Liquid Crystal Display (LCD) phase grating is used for testing convergent optical systems. The rulings are computer-generated and displayed on the LCD. We prove that it is possible to make a variable electronically phase grating by using an LCD. By displaying various phase-shifted rulings and capturing the corresponding ronchigrams, the phase is obtained with the conventional phase-shifting algorithms. Experimental results are shown.

On-Line Signature Verification Based on Genetic Optimization and Neural-Network-Driven Fuzzy Reasoning

This paper presents an innovative approach to solve the on-line signature verification problem in the presence of skilled forgeries. Genetic algorithms (GA) and fuzzy reasoning are the core of our solution. A standard GA is used to find a near optimal representation of the features of a signature to minimize the risk of accepting skilled forgeries. Fuzzy reasoning here is carried out by Neural Networks. The method of a human expert examiner of questioned signatures is adopted here. The solution was tested in the presence of genuine, random and skilled forgeries, with high correct verification rates.

Steps length error detector algorithm in phase-shifting interferometry using Radon transform as a profile measurement

Phase-shifting is one of the most useful methods of phase recovery in digital interferometry in the estimation of small displacements, but miscalibration errors of the phase shifters are very common. In practice, the main problem associated with such errors is related to the response of the phase shifter devices, since they are dependent on mechanical and/or electrical parts. In this work, a novel technique to detect and measure calibration errors in phase-shifting interferometry, when an unexpected phase shift arises, is proposed. The described method uses the Radon transform, first as an automatic-calibrating technique, and then as a profile measuring procedure when analyzing a specific zone of an interferogram. After, once maximum and minimum value parameters have been registered, these can be used to measure calibration errors. Synthetic and real interferograms are included in the testing, which has thrown good approximations for both cases, notwithstanding the interferogram fringe distribution or its phase-shifting steps. Tests have shown that this algorithm is able to measure the deviations of the steps in phase-shifting interferometry. The developed algorithm can also be used as an alternative in the calibration of phase shifter devices.

Optimal Feature Generation with Genetic Algorithms and FLDR in a Restricted-Vocabulary Speech Recognition System

In every pattern recognition problem there exist the need for variable and feature selection and, in many cases, feature generation. In pattern recognition, the term variable is usually understood as the raw measurements or raw values taken from the subjects to be classified, while the term feature is used to refer to the result of the transformations applied to the variables in order to transform them into another domain or space, in which a bigger discriminant capability of the new calculated features is expected; a very popular cases of feature generation are the use of principal component analysis (PCA), in which the variables are projected into a lower dimensional space in which the new features can be used to visualize the underlying class distributions in the original data [1], or the Fourier Transform, in which a few of its coefficients can represent new features [2], [3]. Sometimes, the literature does not make any distinction between variables and features, using them indistinctly [4], [5].