Mario Reyes-Ayala

High-order statistical texture analysis--font recognition applied

A new optical font recognition technique is proposed in this work. The new approach is based on global texture analysis, where statistical methods are used to identify and classify font features. The font recognition is performed by taking the document as a simple image, where one or several types of fonts are present. The identification is not performed letter by letter as with conventional approaches. In the proposed method a window analysis is employed to obtain the features of the document, using fourth and third order moments. The new technique does not involve a study of local typography; therefore, it is content independent. A detailed study was performed with 8 types of fonts commonly used in the Spanish language. Each type of font can have four styles that lead, to 32 font combinations. The font recognition with clean images is 100% accurate. Also, the new method was tested by adding Gaussian noise to clean images, so as to study the impact of image degradation on font recognition. The robustness of the algorithm is also examined in terms of varying resolution.

Inductively-Loaded Yagi-Uda Antenna With Cylindrical Cover for Size Reduction at VHF-UHF Bands

A new technique is employed to reduce the size of Yagi-Uda antennas. The technique consists of adding short circuited cylindrical covers to the structure, making the elements be inductively loaded, and as a result, increase the electric length. A prototype resonating at 660 MHz is developed, and compared to a conventional Yagi-Uda antenna. An effective area reduction of 35% is achieved without decreasing, considerably, the gain, preserving the bandwidth, and front to back lobe ratio.

A Novel Active Load Linearizer for HBT Low Noise Amplifier at 2.4 GHz

A novel active load linearizer is introduced to enhance the non-linear performance of a HBT Low Noise Amplifier. This active load is configured as a Current Source. The results of this structure are compared to an equally biased non-active loaded device, obtaining a higher third order Input Intercept Point (IIP3) around 17 dBm, and a higher compression point (G1 dB) around -1 dBm.

Three Comparative Methods for Mammograms Segmentation: by Master-Slave Algorithm and by Two Parallel Fusion Algorithms

In this paper we compare three different segmentation schemes, in order to segment mammograms (taken from MIAS database): the first one is our proposed Master-Slave method, the second one is the parallel data fusion algorithms such as Bayes formalism and the third one is Dempster-Shafer theory. The Master-Slave algorithm works under the following logic: there are two classifiers (Master or slave), if Master classifier rejects the feature vector (bad classification), then at the same pixel centered-window there are a second features estimation taking a bigger or a smaller window, this second feature vector is now classified by Slave classifier (the decision could be the reject class). Both classifier decisions (Gaussian or KNN) pass to the reconstruction task giving the final mammography segmentation. For the parallel data fusion algorithms we have to estimate at the same time two vector features (one for bigger window and other for smaller window), both feature vector are classified; their decisions could be well or bad classification. From the classifier decisions, we have implemented the Bayes fusion which is based on the Belief function taking the probability to belong a given class. We have implemented also the fusion by Dempster-Shafer theory.