Catálogo de publicaciones - libros

Infrared remote sensing is an extended technique to measure ”in situ” atmospheric pollutant gas concentration. However, retrieval of concentrations from the absorbance spectra provided by technique is not a straightforward problem. In this work the use of artificial neural networks to analyze infrared absorbance spectra is proposed. A summary of classical retrieval codes is presented, highlighting advantages and important drawbacks that arise when these methods are applied to spectral analysis. As an alternative, a neural network retrieval approach is suggested, based on a multi layer perceptron. This approach has been focused to the retrieval of carbon monoxide concentration, because of the great environmental importance of this gas. Absorption overlapping of atmospheric gases such as carbon dioxide, nitrous oxide or water vapour is one the most important problem in the retrieval process. The training dataset has been generated with special care to overcome this aspect and guarantee a successful training phase. Results obtained from the ANN method are very promising. However, high retrieval errors have been found when ANN method is applied to experimental spectra. This fact reveals the need of a deep study of the instrumental parameters to be included in the model.

This paper presents a novel Neural Network application in order to estimate Air Pollutant Levels. The application considers both Pollutant concentrations and Meteorological variables. In order to compute the Air Pollutant Level the method considers three important stages. In first stage, A process to validate data information and built a threedimensional Information Feature Vector with Pollutant concentrations and both wind speed and wind direction meteorological variables is developed. The information Feature Vector is orderly like a time series to estimate the Air Pollutant Level. In second stage, considering the behavior space knowledge a priori about pollutant and meteorological variables distribution a threedimensional Representative Vector is built in order to reduces the computational cost in Neural Network training process. In last stage, a Neural Network is designed and trained with the Threedimensional Representative Vector, then using the Threedimensional Information Feature Vector the Air Pollutant Level is estimated. This paper considers a real time series from an Automatic Environmental Monitoring Network from Salamanca, Guanajuato, Mexico, and therefore in this proposal a real Air Pollutant Level is also estimated.

In this paper the application of an evolutionary procedure based on genetic algorithms for obtaining the dielectric properties of arbitrary shaped, homogeneous or inhomogeneous materials is presented. The optimization procedure matches the measured and simulated scattering parameters of a waveguide setup that contains the sample under study. Depending on the geometry of the sample, analytic or numerical (2D or 3D) electromagnetic simulations must be carried out in order to obtain the simulated scattering parameters for a set of electric permittivities. Results for different polymeric and biological materials are presented with similar uncertainties than conventional direct methods, with the advantage that this new technique can deal with non-canonical and heterogeneous samples.

When a tutoring aims to guide students in the teaching/learning process, it needs to know what knowledge the student has and what goals the student is currently trying to achieve. The Bayesian framework offers a number of techniques for inferring individual’s knowledge state from evidence of mastery of concepts or skills. Using Bayesian networks, we have devised the probabilistic student models for MacBay, a tutoring system that is an authoring tool. MacBay’s models provide prediction of student’s action during teaching/learning process. We combined the Concept Maps and the Bayesian networks in order to obtain a Concept Map with intelligent behavior, where ”intelligence” is considered as the capacity to adapt the interaction to its user’s specific needs. In this paper we describe the way in which we do this combination and inference process.

In this work we implement a neural architecture for recognizing the direction of movement using neural properties that are consistent with biological findings like intrinsic plasticity and synaptic metaplasticity. The network architecture has two memory layers and two competitive layers. This un-supervised neural network is able to identify the direction of movement of an object, being a promising network for object tracking, hand-written and speech recognition.

This contribution consists on the application of a hybrid technique of signals digital processing and artificial intelligence, to classify two kinds of biomedical spectra, normal brain and meningioma tumor. Each signal is processed to extract the relevant information within the range of interest. Then, a Haar 4 wavelet transform is applied to reduce the size of the spectrum without loosing its main features. This signal approximation is coded in a binary set which keeps the frequencies that could have representative amplitude peaks of each signal. The coding is input in a recursive Hamming neural network previously trained, which is able to classify it by comparing it with patterns. The results of the classification are shown for a group of signals that corresponds to human brain tissue. The advantages and disadvantages of the implemented method are discussed.