Showing 1 to 100 of 150 matching Articles
Results per page:
Export (CSV)
By
Gelbukh, Alexander; Sidorov, Grigori
2 Citations
Parallel text alignment is a special type of pattern recognition task aimed to discover the similarity between two sequences of symbols. Given the same text in two different languages, the task is to decide which elements—paragraphs in case of paragraph alignment—in one text are translations of which elements of the other text. One of the applications is training training statistical machine translation algorithms. The task is not trivial unless detailed text understanding can be afforded. In our previous work we have presented a simple technique that relied on bilingual dictionaries but does not perform any syntactic analysis of the texts. In this paper we give a formal definition of the task and present an exact optimization algorithm for finding the best alignment.
more …
By
Oca, Víctor Montes; Torres, Miguel; Levachkine, Serguei; Moreno, Marco
Show all (4)
In this paper, we propose the use of a knowledge based system, which has been implemented in SWIProlog to approach the automatic description of spatial data by means of some logic rules. The process to establish the predicates is based on the topological and geometrical analysis of spatial data. These predicates are handled by a set of rules, which are used to find the relations between geospatial objects. Moreover, the rules aid the searching of several features that compose the partition of topographic maps. For instance, in the case that any road intersects with other, we appreciate that a connection relation exists between different destinies, which can be accessed by these roads. Furthermore, the rules help us to know each possible access for this case. Therefore, this description assists in the tasks of geospatial data interpretation (map description) in order to provide quality information for spatial decision support systems.
more …
By
HernándezRodríguez, Selene; CarrascoOchoa, J. A.; MartínezTrinidad, J. Fco.
The k nearest neighbor (kNN) classifier has been extensively used as a nonparametric technique in Pattern Recognition. However, in some applications where the training set is large, the exhaustive kNN classifier becomes impractical. Therefore, many fast kNN classifiers have been developed to avoid this problem. Most of these classifiers rely on metric properties, usually the triangle inequality, to reduce the number of prototype comparisons. However, in soft sciences, the prototypes are usually described by qualitative and quantitative features (mixed data), and sometimes the comparison function does not satisfy the triangle inequality. Therefore, in this work, a fast k most similar neighbor (kMSN) classifier, which uses a Tree structure and an Approximating and Eliminating approach for Mixed Data, not based on metric properties (Tree AEMD), is introduced. The proposed classifier is compared against other fast kNN classifiers.
more …
By
GonzálezGómez, Efrén; Levachkine, Serguei
Hard problem of cartographic pattern recognition in fine scale maps, using information that comes from coarse scale maps, is considered. The maps are rasterscanned color maps of different thematic, representing the same territory in coarse and fine scale respectively. A solution called CoarsetoFine Scale Method is proposed. This method is defined in terms of means: coarse scale maps and their information; concepts: image associated function, cartographic knowledge domain and cartographic pattern; and tools: a set of clustering criteria of the Logical Combinatorial Pattern Recognition.
more …
By
Batyrshin, Ildar; Sheremetov, Leonid
3 Citations
The methods of pattern recognition in time series based on moving approximation (MAP) transform and MAP image of patterns are proposed. We discuss main properties of MAP transform, introduce a concept of a MAP image of time series and distance between time series patterns based on this concept which were used for recognition of small patterns in noisy time series. To illustrate the application of this technique to recognition of perception based patterns given by sequence of slopes, an example of recognition of water production patterns in petroleum wells used in expert system for diagnosis of water production problems is considered.
more …
By
OlveraLópez, J. Arturo; MartínezTrinidad, J. Francisco; CarrascoOchoa, J. Ariel
1 Citations
In supervised classification, the object selection or instance selection is an important task, mainly for instancebased classifiers since through this process the time in training and classification stages could be reduced. In this work, we propose a new mixed data object selection method based on clustering and border objects. We carried out an experimental comparison between our method and other object selection methods using some mixed data classifiers.
more …
By
GarcíaPerera, L. Paola; MexPerera, Carlos; NolazcoFlores, Juan A.
Abstract
In this research we present a new scheme for the generation of a biometric key based on Automatic Speech Technology and Support Vector Machines. Keys are produced by making a distinction among the voice attributes of the users employing hyperplanes. It is described how the key is conformed and the reliability of the method. We depict an experimental evaluation for different values of the parameters. Among the different kernels for the Support Vector Machine, the RBF obtained the best results.
more …
By
BayroCorrochano, Eduardo
In this paper the authors use the framework of geometric algebra for applications in computer vision, robotics and learning . This mathematical system keeps our intuitions and insight of the geometry of the problem at hand and it helps us to reduce considerably the computational burden of the problems. The authors show that framework of geometric algebra can be in general of great advantage for applications using stereo vision, range data, laser, omnidirectional and odometry based systems. For learning the paper presents the Clifford Support Vector Machines as a generalization of the real and complexvalued Support Vector Machines.
more …
By
Montero, José Antonio; Sucar, L. Enrique
1 Citations
Most gesture recognition systems are based only on hand motion information, and are designed mainly for communicative gestures. However, many activities of everyday life involve interaction with surrounding objects. We propose a new approach for the recognition of manipulative gestures that interact with objects in the environment. The method uses nonintrusive visionbased techniques. The hands of a person are detected and tracked using an adaptive skin color segmentation process, so the system can operate in a wide range of lighting conditions. Gesture recognition is based on hidden Markov models, combining motion and contextual information, where the context refers to the relation of the position of the hand with other objects. The approach was implemented and evaluated on two different domains: video conference and assistance, obtaining gesture recognition rates from 94 % to 99.47 %. The system is very efficient so it is adequate for use in realtime applications.
more …
By
SalgadoGarza, Luis R.; NolazcoFlores, Juan A.
Abstract
This document describes the realization of a spoken information retrieval system and its application to words search in an indexed video database. The system uses an automatic speech recognition (ASR) software to convert the audio signal of a video file into a transcript file and then a document indexing tool to index this transcripted file. Then, a spoken query, uttered by any user, is presented to the ASR to decode the audio signal and propose a hypothesis that is later used to formulate a query to the indexed database. The final outcome of the system is a list of video frame tags containing the audio correspondent to the spoken query. The speech recognition system achieved less than 15% Word Error Rate (WER) and its combined operation with the document indexing system showed outstanding performance with spoken queries.
more …
By
HernándezRodríguez, Selene; MartínezTrinidad, J. Francisco; CarrascoOchoa, J. Ariel
1 Citations
In this work, a fast k most similar neighbor (kMSN) classifier for mixed data is presented. The k nearest neighbor (kNN) classifier has been a widely used nonparametric technique in Pattern Recognition. Many fast kNN classifiers have been developed to be applied on numerical object descriptions, most of them based on metric properties to avoid object comparisons. However, in some sciences as Medicine, Geology, Sociology, etc., objects are usually described by numerical and non numerical features (mixed data). In this case, we can not assume the comparison function satisfies metric properties. Therefore, our classifier is based on search algorithms suitable for mixed data and nonmetric comparison functions. Some experiments and a comparison against other two fast kNN methods, using standard databases, are presented.
more …
By
Calvo, Hiram; Gelbukh, Alexander
1 Citations
We present a method for extracting selectional preferences of verbs from unannotated text. These selectional preferences are linked to an ontology (e.g. the hypernym relations found in WordNet), which allows for extending the coverage for unseen valency fillers. For example, if drink vodka is found in the training corpus, a whole WordNet hierarchy is assigned to the verb todrink (drink liquor, drink alcohol, drink beverage, drink substance, etc.), so that when drink gin is seen in a later stage, it is possible to relate the selectional preference drink vodka with drink gin (as ginis a cohyponym of vodka). This information can be used for word sense disambiguation, prepositional phrase attachment disambiguation, syntactic disambiguation, and other applications within the approach of patternbased statistical methods combined with knowledge. As an example, we present an application to word sense disambiguation based on the Senseval2 training text for Spanish. The results of this experiment are similar to those obtained by Resnik for English.
more …
By
GómezGil, Pilar; delosSantosTorres, Guillermo; RamírezCortés, Manuel
Abstract
When input data is noisy and with a lack of uniformity, classification is a very difficult problem, because decision regions are hard to define in an optimal way. This is the case of recognition of old handwritten manuscript characters, where patterns of the same class may be very different from each other, and patterns of different classes may be similar in terms of Euclidian distances between their feature vectors. In this paper we present the results obtained when a nonsupervised method is used to create feature maps of possible classes in handwriting letters. The prototypes generated in the map present a topological relationship; therefore similar prototypes are near each other. This organization helps to solve the problem of variance in the patterns, allowing a better classification when compared with other supervised classification method, a nearestneighbor algorithm. The feature map was built using a Selforganized Feature Map (SOFM) neural network.
more …
By
Vázquez, Roberto A.; Sossa, Humberto
8 Citations
In this paper we describe how associative memories can be applied to categorize images. If we present to an associative memory (AM) an image we would expect that the AM would respond with something that describes the content of the image; for example, if the image contains a tiger we would expect that the AM would respond with the word “tiger”. In order to achieve this goal, we first chose a set of images. Each image is next associated to the word that better describes the content of the image. With this information an AM is trained as in [10]. We then use the AM to categorize instances of images with the same content even if these images are distorted by some kind of noise. The accuracy of the proposal is tested using a set of images containing different species of flowers and animals.
more …
By
Shin, Kwangcheol; Han, SangYong; Gelbukh, Alexander; Park, Jaehwa
Show all (4)
1 Citations
MEDLINE is a very large database of abstracts of research papers in medical domain, maintained by the National Library of Medicine. Documents in MEDLINE are supplied with manually assigned keywords from a controlled vocabulary called MeSH terms, classified for each document into major MeSH terms describing the main topics of the document and minor MeSH terms giving more details on the document’s topic. To search MEDLINE, we apply a query expansion strategy through automatic relevance feedback, with the following modification: we assign greater weights to the MeSH terms, with different modulation of the major and minor MeSH terms’ weights. With this, we obtain 16% of improvement of the retrieval quality over the best known system.
more …
By
Bolshakov, Igor A.; GaliciaHaro, Sofia N.
Stable coordinate pairs (SCP) like comentarios y sugerencias ‘comments and suggestions’ or sano y salvo ‘safe and sound’ are rather frequent in texts in Spanish, though there are only few thousands of them in language. We characterize SCPs statistically by a numerical Stable Connection Index and reveal its unimodal distribution. We also propose lexical, morphologic, syntactic, and semantic categories for SCP structural description — for both a whole SCP and its components. It is argued that database containing a set of categorized SCPs facilitates several tasks of automatic NLP.. The research is based on a set of ca. 2200 Spanish coordinate pairs.
more …
By
CoyotlMorales, Rosa María; VillaseñorPineda, Luis; MontesyGómez, Manuel; Rosso, Paolo
Show all (4)
20 Citations
Authorship attribution is the task of identifying the author of a given text. The main concern of this task is to define an appropriate characterization of documents that captures the writing style of authors. This paper proposes a new method for authorship attribution supported on the idea that a proper identification of authors must consider both stylistic and topic features of texts. This method characterizes documents by a set of word sequences that combine functional and content words. The experimental results on poem classification demonstrated that this method outperforms most current stateoftheart approaches, and that it is appropriate to handle the attribution of short documents.
more …
By
Díaz de León, Rocío; Sucar, Luis Enrique
We propose a general model for visual recognition of human activities, based on a probabilistic graphical framework. The motion of each limb and the coordination between them is considered in a layered network that can represent and recognize a wide range of human activities. By using this model and a sliding window, we can recognize simultaneous activities in a continuous way. We explore two inference methods for obtaining the most probable set of activities per window: probability propagation and abduction. In contrast with the standard approach that uses several models, we use a single classifier for multiple activity recognition. We evaluated the model with real image sequences of 6 different activities performed continuously by different people. The experiments show high recall and recognition rates.
more …
By
Sossa, Humberto; Barrón, Ricardo; Cuevas, Francisco; Aguilar, Carlos; Cortés, Héctor
Show all (5)
We show how the binary αβ associative memories recently proposed by Yáñez in [1] can be extended to work now in the graylevel case. To get the desired extension we take the operators α and β, foundation of the αβ memories, and propose a more general family of operators among them the original operators α and β are a subset. For this we formulate a set of functional equations, solve this system and find a family of solutions. We show that the α and β originally proposed in [1] are just a particular case of this new family. We give the properties of the new operators. We then use these operators to build the extended memories. We provide the conditions under which the proposed extended memories are able to recall a pattern either from the pattern’s fundamental set or from altered versions of them. We provide real examples with images where the proposed memories show their efficiency.
more …
By
BayroCorrochano, Eduardo
Abstract
In this paper the authors use the framework of geometric algebra for applications in computer vision, robotics and learning . This mathematical system keeps our intuitions and insight of the geometry of the problem at hand and it helps us to reduce considerably the computational burden of the problems. The authors show that framework of geometric algebra can be in general of great advantage for applications using stereo vision, range data, laser, omnidirectional and odometry based systems. For learning the paper presents the Clifford Support Vector Machines as a generalization of the real and complexvalued Support Vector Machines.
more …
By
BayroCorrochano, Eduardo; Torre Gomora, Miguel Angel
Abstract
The contribution of this work is to generalize the real and complex wavelet transforms and to derive for the first time a quaternionic wavelet pyramid for multiresolution analysis using three phases. The paper can be very useful for researchers and practitioners interested in understanding and applications of the quaternion wavelet transform.
more …
By
Gelbukh, Alexander; Alexandrov, Mikhail; Han, SangYong
6 Citations
One of the most important steps in text processing and information retrieval is stemming – reducing of words to stems expressing their base meaning, e.g., bake, baked, bakes, baking → bak. We suggest an unsupervised method of recognition such inflection patterns automatically, with no a priori information on the given language, basing exclusively on a list of words extracted from a large text. For a given word list V we construct two sets of strings: stems S and endings E, such that each word from V is a concatenation of a stem from S and ending from E. To select an optimal model, we minimize the total number of elements in S and E. Though such a simplistic model does not reflect many phenomena of real natural language morphology, it shows surprisingly promising results on different European languages. In addition to practical value, we believe that this can also shed light on the nature of human language.
more …
By
Carrera, Gerardo; Savage, Jesus; MayolCuevas, Walterio
1 Citations
This paper presents a robust implementation of an object tracker able to tolerate partial occlusions, rotation and scale for a variety of different objects. The objects are represented by collections of interest points which are described in a multiresolution framework, giving a representation of those points at different scales. Inspired by [1], a stack of descriptors is built only the first time that the interest points are detected and extracted from the region of interest. This provides efficiency of representation and results in faster tracking due to the fact that it can be done offline. An Unscented Kalman Filter (UKF) using a constant velocity model estimates the position and the scale of the object, with the uncertainty in the position and the scale obtained by the UKF, the search of the object can be constrained only in a specific region in both the image and in scale.
The use of this approach shows an improvement in realtime tracking and in the ability to recover from full occlusions.
more …
By
Sossa, Humberto; Barrón, Ricardo; Vázquez, Roberto A.
10 Citations
In this note we describe a new set of associative memories able to recall patterns in the presence of mixed noise. Conditions are given under which the proposed memories are able to recall patterns either from the fundamental set of patterns and from distorted versions of them. Numerical and real examples are also provided to show the efficiency of the proposal.
more …
By
LópezEspinoza, Erika; CarrascoOchoa, Jesús Ariel; MartínezTrinidad, José Fco.
In this paper, two strategies to compute the support sets system for the supervised classifier ALVOT (voting algorithms) using sequential floating selection are presented. ALVOT is a supervised classification model based on the partial precedence principle, therefore, it needs, as feature selection, a set of features subsets, this set is called support sets system. The sequential floating selection methods for feature selection find only one relevant features subset. The introduced strategies search for a set of features subsets to generate a support sets system. Both strategies are compared between them and against the feature selection method based on testor theory, which is commonly used to compute this system. Results obtained with both strategies on different databases from UCI and on the faces database from Olivetti Research Laboratory (ORL) in Cambridge are presented.
more …
By
OlveraLópez, J. Arturo; MartínezTrinidad, J. Fco.; CarrascoOchoa, J. Ariel
4 Citations
Edition is an important and useful task in supervised classification specifically for instancebased classifiers because edition discards from the training set those useless or harmful objects for the classification accuracy and it helps to reduce the size of the original training sample and to increase both the classification speed and accuracy. In this paper, we propose two edition schemes that combine edition methods and sequential search for instance selection. In addition, we present an empirical comparison between these schemes and some other edition methods.
more …
By
GómezGil, Pilar; delosSantosTorres, Guillermo; RamírezCortés, Manuel
1 Citations
When input data is noisy and with a lack of uniformity, classification is a very difficult problem, because decision regions are hard to define in an optimal way. This is the case of recognition of old handwritten manuscript characters, where patterns of the same class may be very different from each other, and patterns of different classes may be similar in terms of Euclidian distances between their feature vectors. In this paper we present the results obtained when a nonsupervised method is used to create feature maps of possible classes in handwriting letters. The prototypes generated in the map present a topological relationship; therefore similar prototypes are near each other. This organization helps to solve the problem of variance in the patterns, allowing a better classification when compared with other supervised classification method, a nearestneighbor algorithm. The feature map was built using a Selforganized Feature Map (SOFM) neural network.
more …
By
LópezEspinoza, Erika; CarrascoOchoa, Jesús Ariel; MartínezTrinidad, José Fco.
Abstract
In this paper, two strategies to compute the support sets system for the supervised classifier ALVOT (voting algorithms) using sequential floating selection are presented. ALVOT is a supervised classification model based on the partial precedence principle, therefore, it needs, as feature selection, a set of features subsets, this set is called support sets system. The sequential floating selection methods for feature selection find only one relevant features subset. The introduced strategies search for a set of features subsets to generate a support sets system. Both strategies are compared between them and against the feature selection method based on testor theory, which is commonly used to compute this system. Results obtained with both strategies on different databases from UCI and on the faces database from Olivetti Research Laboratory (ORL) in Cambridge are presented.
more …
By
CarrascoOchoa, Jesús Ariel; RuizShulcloper, José; DelaVegaDoría, Lucía Angélica
Typical e:testors are useful to do feature selection in supervised classification problems with mixed incomplete data, where similarity function is not the total coincidence, but it is a one threshold function. In this kind of problems, modifications on the training matrix can appear very frequently. Any modification of the training matrix can change the set of all typical ε:testors, so this set must be recomputed after each modification. But, complexity of algorithms for calculating all typical ε:testors of a training matrix is too high. In this paper we analyze how the set of all typical ε:testors changes after modifications. An alternative method to calculate all typical ε:testors of the modified training matrix is exposed. The new method’s complexity is analyzed and some experimental results are shown.
more …
By
Díaz de León, Juan Luis; Yánez, C.; Guzmán, Giovanni
This paper presents a novel algorithm intended to generate kconnected skeletons of a digital binary image using a new mask set. These skeletons may be 4 or 8 connected. The new algorithm performs a thinning process that finish when it is not possible to eliminate additional pixels without breaking the connectivity. The endpoint criterion and a 3x3 masks set are used to decide if a pixel is eliminated. The proposed masks set for each kind of connectivity covers all the necessary cases, and guarantee to obtain a one pixel wide and kconnected skeleton without parasitic branches. The new algorithm yields some advantages to developers. It is not just oriented to written characters or some kind of object in particular; this means that the algorithm can be adapted easily to any application generating good results. Besides, the user can work with different classes of connectivity; note that several recent algorithms use 4connectivity while 8connectivity is used for others. Additionally, the skeletons produced by the new algorithm are immune to structured noise around the processed objects.
more …
By
Cumplido, René; CarrascoOchoa, J. Ariel; Feregrino, Claudia
4 Citations
Typical testors are a useful tool to do feature selection in supervised classification problems with mixed incomplete data. However, the complexity of computing all typical testors of a training matrix has an exponential growth with respect to the number of columns in the matrix. For this reason different approaches like heuristic algorithms, parallel and distributed processing, have been developed. In this paper, we present a configurable custom architecture for the efficient identification of testors from a given input matrix. The architectural design is based on a brute force approach that is suitable for high populated input matrixes. The architecture has been designed to deal with parallel processing and can be configured for any size of matrix. The architecture is able to evaluate if a vector is a testor of the matrix in a single clock cycle. The architecture has been implemented on a Field Programmable Gate Array (FPGA) device. Results show that it provides runtime improvements over software implementations running on stateoftheart processors. FPGA implementation results are presented and implications to the field of pattern recognition discussed.
more …
By
AyaquicaMartínez, I. O.; MartínezTrinidad, J. Fco.; CarrascoOchoa, J. Ariel
The kmeans algorithm is the most studied and used tool for solving the clustering problem when the number of clusters is known a priori. Nowadays, there is only one conceptual version of this algorithm, the conceptual kmeans algorithm. One characteristic of this algorithm is the use of generalization lattices, which define relationships among the feature values. However, for many applications, it is difficult to determine the best generalization lattices; moreover there are not automatic methods to build the lattices, thus this task must be done by the specialist of the area in which we want to solve the problem. In addition, this algorithm does not work with missing data. For these reasons, in this paper, a new conceptual kmeans algorithm that does not use generalization lattices to build the concepts and allows working with missing data is proposed. We use complex features for generating the concepts. The complex features are subsets of features with associated values that characterize objects of a cluster and at the same time not characterize objects from other clusters. Some experimental results obtained by our algorithm are shown and they are compared against the results obtained by the conceptual kmeans algorithm.
more …
By
Altamirano, Luis Carlos; Altamirano Robles, Leopoldo; Alvarado, Matías
In this work, a generalization of nonuniform sampling technique to construct appearancebased models is proposed. This technique analyses the object appearance defined by several parameters of variability, determining how many and which images are required to model appearance, with a given precision ε. Throughout nonuniform sampling, we obtain a guideline to spend less time on model construction and to diminish storage, when pose estimation no matters. The proposed technique is based on a scheme of Nlinear interpolation and SSD (SumofSquaredDifference) distance, and it is used in conjunction with the eigenspaces method for object recognition. Experimental results showing the advantages are exposed.
more …
By
BayroCorrochano, Eduardo; ZamoraEsquivel, Julio; LópezFranco, Carlos
In this paper the authors use the framework of conformal geometric algebra for the treatment of robot vision tasks. In this mathematical system we calculated projective invariants using omnidirectional vision for object recognition. We show the power of the mathematical system for handling differential kinematics in visual guided tracking.
more …
By
GarcíaPerera, L. Paola; MexPerera, Carlos; NolazcoFlores, Juan A.
2 Citations
In this research we present a new scheme for the generation of a biometric key based on Automatic Speech Technology and Support Vector Machines. Keys are produced by making a distinction among the voice attributes of the users employing hyperplanes. It is described how the key is conformed and the reliability of the method. We depict an experimental evaluation for different values of the parameters. Among the different kernels for the Support Vector Machine, the RBF obtained the best results.
more …
By
BayroCorrochano, Eduardo; Torre Gomora, Miguel Angel
3 Citations
The contribution of this work is to generalize the real and complex wavelet transforms and to derive for the first time a quaternionic wavelet pyramid for multiresolution analysis using three phases. The paper can be very useful for researchers and practitioners interested in understanding and applications of the quaternion wavelet transform.
more …
By
Moreno, Marco; Levachkine, Serguei; Torres, Miguel; Quintero, Rolando
Show all (4)
4 Citations
We present an approach to perform a landform classification of raster geoimages to obtain the semantics of DEMs. We consider the following raster layers: slope, profile curvature and plan curvature, which have been built to identify the intrinsic properties of the landscape. We use a multivalued raster to integrate these layers. The attributes of the multivalued raster are classified to identify the landform elements. The classification approach is used to find the terrain characteristics of the water movement. Moreover, we describe the mechanisms to compute the primary attributes of digital terrain model. The method has been implemented into Geographical Information System–ArcInfo, and applied for Tamaulipas State, Mexico.
more …
By
GarcíaHernández, René A.; MartínezTrinidad, José Fco.; CarrascoOchoa, Jesús Ariel
9 Citations
One of the sequential pattern mining problems is to find the maximal frequent sequences in a database with a β support. In this paper, we propose a new algorithm to find all the maximal frequent sequences in a text instead of a database. Our algorithm in comparison with the typical sequential pattern mining algorithms avoids the joining, pruning and text scanning steps. Some experiments have shown that it is possible to get all the maximal frequent sequences in a few seconds for medium texts.
more …
By
HernándezReyes, Edith; MartínezTrinidad, J. Fco.; CarrascoOchoa, J. A.; GarcíaHernández, René A.
Show all (4)
In document clustering, documents are commonly represented through the vector space model as a word vector where the features correspond to the words of the documents. However, there are a lot of words in a document set; therefore the vector size could be enormous. Also, the vector space model does not take into account the word order that could be useful to group similar documents. In order to reduce these disadvantages, we propose a new document representation in which each document is represented as a set of its maximal frequent sequences. The proposed document representation is applied for document clustering and the quality of the clustering is evaluated through internal and external measures, the results are compared with those obtained with the vector space model.
more …
By
SalgadoGarza, Luis R.; NolazcoFlores, Juan A.
This document describes the realization of a spoken information retrieval system and its application to words search in an indexed video database. The system uses an automatic speech recognition (ASR) software to convert the audio signal of a video file into a transcript file and then a document indexing tool to index this transcripted file. Then, a spoken query, uttered by any user, is presented to the ASR to decode the audio signal and propose a hypothesis that is later used to formulate a query to the indexed database. The final outcome of the system is a list of video frame tags containing the audio correspondent to the spoken query. The speech recognition system achieved less than 15% Word Error Rate (WER) and its combined operation with the document indexing system showed outstanding performance with spoken queries.
more …
By
Shin, Kwangcheol; Han, SangYong; Gelbukh, Alexander; Park, Jaehwa
Show all (4)
Abstract
MEDLINE is a very large database of abstracts of research papers in medical domain, maintained by the National Library of Medicine. Documents in MEDLINE are supplied with manually assigned keywords from a controlled vocabulary called MeSH terms, classified for each document into major MeSH terms describing the main topics of the document and minor MeSH terms giving more details on the document’s topic. To search MEDLINE, we apply a query expansion strategy through automatic relevance feedback, with the following modification: we assign greater weights to the MeSH terms, with different modulation of the major and minor MeSH terms’ weights. With this, we obtain 16% of improvement of the retrieval quality over the best known system.
more …
By
Ponomaryov, Volodymyr; GallegosFunes, Francisco; RosalesSilva, Alberto; Loboda, Igor
Show all (4)
1 Citations
We present novel algorithms to suppress impulsive noise in video color sequences. They use order statistics, directional and adaptive processing techniques. Extensive simulation results in known reference video color sequences have demonstrated that the proposed filters consistently outperform other filters by balancing the tradeoff between noise suppression, detail preservation, and chromaticity characteristics. The criteria used to compare the performance or various filters were the PSNR, MAE, and NCD.
more …
By
GodoyCalderón, Salvador; MartínezTrinidad, J. Fco.; Cortés, Manuel Lazo
There is presently no unified methodology that allows the evaluation of supervised and nonsupervised classification algorithms. Supervised problems are evaluated through Quality Functions that require a previously known solution for the problem, while nonsupervised problems are evaluated through several Structural Indexes that do not evaluate the classification algorithm by using the same pattern similarity criteria embedded in the classification algorithm. In both cases, a lot of useful information remains hidden or is not considered by the evaluation method, such as the quality of the supervision sample or the structural change generated by the classification algorithm on the sample. This paper proposes a unified methodology to evaluate classification problems of both kinds, that offers the possibility of making comparative evaluations and yields a larger amount of information to the evaluator about the quality of the initial sample, when it exists, and regarding the change produced by the classification algorithm.
more …
By
RamosMichel, Erika M.; Kober, Vitaly
2 Citations
Several correlation filters are derived to improve pattern recognition of a noisy target embedded into nonoverlapping background, when the input image is degraded with a linear system. With the help of computer simulation we analyze and compare the performance of various correlation based methods for reliable detection and localization of objects in blurred and noisy images.
more …
By
Sossa, Humberto; Flusser, Jan
Geometric moments have been proven to be a very efficient tool for description and recognition of binary shapes. Numerous methods for effective calculation of image moments have been presented up to now. Recently, Sossa, Yañez and Díaz [Pattern Recognition, 34(2):271276, 2001] proposed a new algorithm based on a morphologic decomposition of the image into a set of closed disks. Their algorithm yields approximative results. In this paper we propose a refinement of their method that performs as fast as the original one but gives exact results.
more …
By
Flores, Norberto; KuriMorales, Ángel; Gamio, Carlos
1 Citations
The article presents a possible solution to a typical tomographic images generation problem from data of an industrial process located in a pipeline or vessel. These data are capacitance measurements obtained noninvasively according to the well known ECT technique (Electrical Capacitance Tomography). Every 313 pixels image frame is derived from 66 capacitance measurements sampled from the real time process. The neural nets have been trained using the backpropagation algorithm where training samples have been created synthetically from a computational model of the real ECT sensor. To create the image 313 neuronal nets, each with 66 inputs and one output, are used in parallel. The resulting image is finally filtered and displayed. The different ECT system stages along with the different tests performed with synthetic and real data are reported. We show that the image resulting from our method is a faster and more precise practical alternative to previously reported ones.
more …
By
Alejo, R.; Sotoca, J. M.; Casañ, G. A.
4 Citations
The latest research in neural networks demonstrates that the class imbalance problem is a critical factor in the classifiers performance when working with multiclass datasets. This occurs when the number of samples of some classes is much smaller compared to other classes. In this work, four different options to reduce the influence of the class imbalance problem in the neural networks are studied. These options consist of introducing several cost functions in the learning algorithm in order to improve the generalization ability of the networks and speed up the convergence process.
more …
By
RiveraRovelo, Jorge; Herold, Silena; BayroCorrochano, Eduardo
2 Citations
In this paper we present a method based on selforganizing neural networks to extract the shape of a 2D or 3D object using a set of transformations expressed as versors in the conformal geometric algebra framework. Such transformations, when applied to any geometric entity of this geometric algebra, define the shape of the object. This approach was tested with several images, but here we show its utility first using a 2D magnetic resonance image to segment the ventricle. Then we present some examples of an application for the case of 3D objects.
more …
By
Lira, Jorge
2 Citations
A model of desertification in semiarid environment employing satellite multispectral images is presented. The variables proposed to characterize desertification are: texture of terrain, vegetation index for semiarid terrain, and albedo of terrain. The texture is derived from a divergence operator applied upon the vector field formed by the first three principal components of the image. The vegetation index selected is the TSAVI, suitable for semiarid environment where vegetation is scarce. The albedo is calculated from the first principal component obtained from the bands of the multispectral image. These three variables are input into a clustering algorithm resulting in six desertification grades. These grades are ordered from nodesertification to severe desertification. Details are provided for the computer calculation of the desertification variables, and the parameters employed in the clustering algorithm. A multispectral Landsat TM image is selected for this research. A thematic map of desertification is then generated with the support of ancillary data related to the study area.
more …
By
Yáñez, Cornelio; FelipeRiveron, Edgardo; LópezYáñez, I.; FloresCarapia, R.
Show all (4)
6 Citations
In this paper the design and operation of an Automatic Color Matching system is presented. This novel system takes advantage of the improvements introduced by AlphaBeta associative memories, an efficient, unconventional model of associative memory of recent creation. The results are demonstrated through experiments on a relatively small database with 1001 samples prepared by the authors. However, the approach is considered valid according to the tendency of the results obtained, in part, thanks to the performance exhibited by AlphaBeta associative memories.
more …
By
Urcid, Gonzalo; Ritter, Gerhard X.; Iancu, Laurentiu
5 Citations
Morphological perceptrons use a lattice algebra approach to learn and classify a set of patterns. Dendritic structure combined with lattice algebra operations have properties that are completely different than those of traditional perceptron models. In the present paper, we focus our attention in single layer morphological perceptrons that classify correctly the parity of all bit strings of length n, as a oneclass pattern recognition problem. The nbit parity problem is the ndimensional extension of the classic XOR problem in the Euclidean plane and is commonly used as a difficult benchmark to test the performance of training algorithms in artificial neural networks. We present results for values of n up to 10, obtained with a training algorithm based on elimination.
more …
By
MarinCastro, Heidy; Sucar, Enrique; Morales, Eduardo
4 Citations
Automatic image annotation consists on automatically labeling images, or image regions, with a predefined set of keywords, which are regarded as descriptors of the highlevel semantics of the image. In supervised learning, a set of previously annotated images is required to train a classifier. Annotating a large quantity of images by hand is a tedious and time consuming process; so an alternative approach is to label manually a small subset of images, using the other ones under a semisupervised approach. In this paper, a new semisupervised ensemble of classifiers, called WSA, for automatic image annotation is proposed. WSA uses naive Bayes as its base classifier. A set of these is combined in a cascade based on the AdaBoost technique. However, when training the ensemble of Bayesian classifiers, it also considers the unlabeled images on each stage. These are annotated based on the classifier from the previous stage, and then used to train the next classifier. The unlabeled instances are weighted according to a confidence measure based on their predicted probability value; while the labeled instances are weighted according to the classifier error, as in standard AdaBoost. WSA has been evaluated with benchmark data sets, and 2 sets of images, with promising results.
more …
By
LópezEspinoza, Erika Danaé; AltamiranoRobles, Leopoldo
In this paper, the deterministic component of 2D Wold decomposition is used to obtain texture descriptors in industrial plastic quality images, and hidden geometry of tree crown in remote sensing images. The texture image is decomposed into two texture images: a nondeterministic texture and a deterministic one. In order to obtain texture descriptors, a set of discriminant texture features is selected from the deterministic component. The texture descriptors have been used to distinguish among three kinds of plastic quality. The obtained texture descriptors are compared against texture descriptors obtained from the original image. With the objective to find hidden geometry of tree crown in remote sensing images, the deterministic component of the original image is analyzed. The observed geometry is compared against the modeled geometry in the literature of marked point processes.
more …
By
Starostenko, Oleg; ChávezAragón, Alberto; Sánchez, J. Alfredo; Ostróvskaya, Yulia
Show all (4)
This paper presents a novel approach for image retrieval from digital collections. Specifically, we describe IRONS (Image Retrieval with Ontological Descriptions of Shapes), a system based on the application of several novel algorithms that combine lowlevel image analysis techniques with automatic shape extraction and indexing. In order to speed up preprocessing, we have proposed and implemented the convex regions algorithm and discrete curve evolution approach. The image indexing module of IRONS is addressed using two proposed algorithms: the tangent space and the twosegment turning function for shapes representation invariant to rotation and scale. Another goal of the proposed method is the integration of useroriented descriptions, which leads to more complete retrieval by accelerating the convergence to the expected result. For the definition of image semantics, ontology annotation of subregions has been used.
more …
By
AyaquicaMartínez, I. O.; MartínezTrinidad, J. F.; CarrascoOchoa, J. A.
The conceptual kmeans algorithm consists of two steps. In the first step the clusters are obtained (aggregation step) and in the second one the concepts or properties for those clusters are generated (characterization step). We consider the conceptual kmeans management of mixed, qualitative and quantitative, features is inappropriate. Therefore, in this paper, a new conceptual kmeans algorithm using similarity functions is proposed. In the aggregation step we propose to use a different clustering strategy, which allows working in a more natural way with object descriptions in terms of quantitative and qualitative features. In addition, an improvement of the characterization step and a new quality measure for the generated concepts are presented. Some results obtained after applying both, the original and the modified algorithms on different databases are shown. Also, they are compared using the proposed quality measure.
more …
By
HidalgoSilva, Hugo
A Gabor based representation for textured images is proposed. Instead of the ordinary filter bank, a reproducing kernel representation is constructed consisting of a sum of several local reproducing kernels. The image representation coefficients are computed by a basis pursuit procedure, and are then considered as the feature vectors. The feature vectors are used to construct a kernel for a support vector classifier. Results are presented for a set of oriented texture images.
more …
By
Gómez, Octavio; González, Jesús A.; Morales, Eduardo F.
6 Citations
Segmentation through seeded region growing is widely used because it is fast, robust and free of tuning parameters. However, the seeded region growing algorithm requires an automatic seed generator, and has problems to label unconnected pixels (the unconnected pixel problem). This paper introduces a new automatic seeded region growing algorithm called ASRGIB1 that performs the segmentation of color (RGB) and multispectral images. The seeds are automatically generated via histogram analysis; the histogram of each band is analyzed to obtain intervals of representative pixel values. An image pixel is considered a seed if its gray values for each band fall in some representative interval. After that, our new seeded region growing algorithm is applied to segment the image. This algorithm uses instancebased learning as distance criteria. Finally, according to the user needs, the regions are merged using ownership tables. The algorithm was tested on several leukemia medical images showing good results.
more …
By
Quintero, Rolando; Levachkine, Serguei; Torres, Miguel; Moreno, Marco
Show all (4)
1 Citations
In this work, we present an algorithm for increasing spectral resolution in DEM. The algorithm is based on the 8connected skeleton of polygons formed by the contour lines and to prune this skeleton by translating it into a graph. This is an alternative to the processes of vector interpolation. With this approach, it is possible to find new elevation data from the information contained in DEM and generate new data with the same spatial resolution.
more …
By
Lara, Carlos; Romero, Leonardo
This paper deals with the problem of finding the movement of a mobile robot given two consecutive laser scans. The proposed method extracts a line map from the sequence of points in each laser scan, using a probabilistic approach, and then computes virtual corners between two lines in the same line map. The movement of the robot is estimated from correspondences of virtual corners between the two line maps. The combination of the probabilistic approach to find lines and the reduced number of virtual corners are the key ideas to get a simple, fast, robust to outliers, and reliable method to solve the local localization problem.
more …
By
Calvo, Hiram; Gelbukh, Alexander
Abstract
We present a method for extracting selectional preferences of verbs from unannotated text. These selectional preferences are linked to an ontology (e.g. the hypernym relations found in WordNet), which allows for extending the coverage for unseen valency fillers. For example, if drink vodka is found in the training corpus, a whole WordNet hierarchy is assigned to the verb todrink (drink liquor, drink alcohol, drink beverage, drink substance, etc.), so that when drink gin is seen in a later stage, it is possible to relate the selectional preference drink vodka with drink gin (as ginis a cohyponym of vodka). This information can be used for word sense disambiguation, prepositional phrase attachment disambiguation, syntactic disambiguation, and other applications within the approach of patternbased statistical methods combined with knowledge. As an example, we present an application to word sense disambiguation based on the Senseval2 training text for Spanish. The results of this experiment are similar to those obtained by Resnik for English.
more …
By
Riverón, Edgardo M. Felipe; Toro Céspedes, Mijail
3 Citations
The objective of this paper is to measure some important parameters of the optic disk (or optic papilla) in ophthalmoscopic color images of human retinas. The approach consists of locating the optic disk automatically, segmenting its contour and the contour of the depressionlike feature caused by glaucoma, called an excavation or cup. Then the corresponding areas are measured to calculate the ratio Cup/Disc and the relative displacement of the centroids of both regions. To achieve these objectives, noise is filtered, luminance is normalized, and a thresholding technique is used. The results obtained will aid the work of ophthalmologists by increasing the quality of automatic diagnosis of glaucoma, one of the main causes of blindness worldwide.
more …
By
Romero, Leonardo; Lara, Carlos
1 Citations
In this work we consider a mobile robot with a laser range finder. Our goal is to find the best set of lines from the sequence of points given by a laser scan. We propose a probabilistic method to deal with noisy laser scans, in which the noise is not properly modeled using a Gaussian Distribution. An experimental comparison with a very well known method (SMSM), using a mobile robot simulator and a real mobile robot, shows the robustness of the new method. The new method is also fast enough to be used in real time.
more …
By
CarrascoOchoa, Jesús Ariel; RuizShulcloper, José; DelaVegaDoría, Lucía Angélica
Abstract
Typical e:testors are useful to do feature selection in supervised classification problems with mixed incomplete data, where similarity function is not the total coincidence, but it is a one threshold function. In this kind of problems, modifications on the training matrix can appear very frequently. Any modification of the training matrix can change the set of all typical ε:testors, so this set must be recomputed after each modification. But, complexity of algorithms for calculating all typical ε:testors of a training matrix is too high. In this paper we analyze how the set of all typical ε:testors changes after modifications. An alternative method to calculate all typical ε:testors of the modified training matrix is exposed. The new method’s complexity is analyzed and some experimental results are shown.
more …
By
Barrón, Ricardo; Sossa, Humberto; Cruz, Benjamín
In this work we present an algorithm for training an associative memory based on the socalled multilayered morphological perceptron with maximal support neighborhoods. We compare the proposal with the original one by performing some experiments with real images. We show the superiority of the new one. We also give formal conditions for correct classification. We show that the proposal can be applied to the case of graylevel images and not only binary images.
more …
By
SuasteRivas, Israel; ReyesGalaviz, Orion F.; DiazMendez, Alejandro; ReyesGarcia, Carlos A.
Show all (4)
8 Citations
In this paper we describe the implementation of a fuzzy relational neural network model. In the model, the input features are represented by fuzzy membership, the weights are described in terms of fuzzy relations. The output values are obtained with the maxmin composition, and are given in terms of fuzzy class membership values. The learning algorithm is a modified version of backpropagation. The system is tested on an infant cry classification problem, in which the objective is to identify pathologies in recently born babies.
more …
By
Díaz, Carlos; Altamirano Robles, Leopoldo
3 Citations
Various algorithms based on unwrapping first the mostreliable pixels have been proposed. These were restricted to continuous path and were subject to troubles on defining an initial pixel. The technique proposed uses a reliability function that helps us to define starting points, it does not follow a continuous path to perform the unwrapping operation, and it uses a mask to pick out invalid pixels. The technique is explained with all the specifics and exemplify with some examples.
more …
By
PerezGarcia, Arturo; AyalaRamirez, Victor; SanchezYanez, Raul E.; AvinaCervantes, JuanGabriel
Show all (4)
4 Citations
In this work, we present a Monte Carlo approach to compute Hausdorff distance for locating objects in real images. Objects are considered to be only under translation motion. We use edge points as the features of the model. Using a different interpretation of the Hausdorff distance, we show how image similarity can be measured by using a randomly subsampled set of feature points. As a result of computing the Hausdorff distance on smaller sets of features, our approach is faster than the classical one. We have found that our method converges toward the actual Hausdorff distance by using less than 20 % of the feature points. We show the behavior of our method for several fractions of feature points used to compute Hausdorff distance. These tests let us conclude that performance is only critically degraded when the subsampled set has a cardinality under 15 % of the total feature points in real images.
more …
By
RomeroMoreno, M.; MartínezTrinidad, J. Fco.; CarrascoOchoa, J. A.
1 Citations
Gait Recognition is a noninvasive biometric technique for identifying persons through the way they walk. Currently there are many Gait Recognition methods, most of them based on a similarity function. In this paper, we propose two new methods for Gait Recognition based on silhouette and contour, using a classifier ensemble. Experimental results on a public standard database are shown and compared against others Gait Recognition methods.
more …
By
Cano, Sergio; Suaste, Israel; Escobedo, Daniel; ReyesGarcía, Carlos A.; Ekkel, Taco
Show all (5)
5 Citations
The present work proposes a combined classifier of infant cry units that links in a single structure two focuses: a thresholdbased classification and ANNbased classification. The thresholdbased classifier considers 4 new acoustic features:stridor, melody, voicedness, shifts, that show properly their robustness in front of alterations of the acoustics of infant cry concerned with the presence of some diseases. In order to satisfy the automatic estimation their practical implementations are also considered. The ANNbased classifier consists in a feedforward network using the method of Scale Gradient Conjugate (MSGC) as learning algorithm and the MFCCs as input vectors to the net. Each focus or classification stage gives in the exit one indicator (FN1 and FN2) that generates to the output a decision on two classes with gradation (normal, moderatelypathologic and pathologic). The results demonstrate the potentiality of these types of combined classifiers when the advantages of each focus in particular are properly emphasized
more …
By
Mozerov, Mikhail; Kober, Vitaly
A new robust matching algorithm for motion detection and computation of precise estimates of motion vectors of moving objects in a sequence of images is presented. Common matching algorithms of dynamic image analysis usually utilize local smoothness constraints. The proposed method exploits global motion smoothness. The suggested matching algorithm is robust to motion discontinuity as well as to noise degradation of a signal. Computer simulation and experimental results demonstrate an excellent performance of the method in terms of dynamic motion analysis.
more …
By
RosalesSilva, Alberto; Ponomaryov, Volodymyr I.; GallegosFunes, Francisco J.
1 Citations
We propose a fuzzy logic recursive scheme using directional processing for motion detection and spatialtemporal filtering to decrease Gaussian noise corruption. We introduce novel ideas that employ the differences between images. That permits to connect these using angle deviations in them obtaining several parameters and applying them in the robust algorithm that is capable to detect and differentiate movement in background of noise in any way.
more …
By
OlveraLópez, J. Arturo; CarrascoOchoa, J. Ariel; MartínezTrinidad, J. Fco.
6 Citations
In Pattern recognition, the supervised classifiers use a training set T for classifying new prototypes. In practice, not all information in T is useful for classification therefore it is necessary to discard irrelevant prototypes from T. This process is known as prototype selection, which is an important task for classifiers since through this process the time in the training and/or classification stages could be reduced. Several prototype selection methods have been proposed following the Nearest Neighbor (NN) rule; in this work, we propose a new prototype selection method based on the prototype relevance and border prototypes, which is faster (over large datasets) than the other tested prototype selection methods. We report experimental results showing the effectiveness of our method and compare accuracy and runtimes against other prototype selection methods.
more …
By
Makagonov, Pavel; Ruiz Figueroa, Alejandro
1 Citations
The problem of measuring and predicting the future of various branches of science is discussed. We propose an economical approach that is useful for the estimation of the stage of development for any branch of “normal” science with the help of abstract flow analysis. For this goal it is necessary to collect large amounts of abstracts uniformly distributed in years. As abstracts are poor knowledge objects, we use the procedure of aggregation in its annual sum of texts as an elemental unit for cluster analysis. For cluster analysis we use the tool kit “Visual Heuristic Cluster Analysis for Texts” developed earlier by one of the coauthors, with K. Sboychakov. To determine the topic of the cluster, we propose to use chapters of manuals and articles principal in the procedure of pattern recognition.
more …
By
MorenoDaniel, Antonio; Juang, BiingHwang; NolazcoFlores, Juan A.
2 Citations
The implementation of a pseudo textindependent Speaker Verification system is described. This system was designed to use only information extracted directly from the coded parameters embedded in the ITUT G.729 bitstream. Experiments were performed over the YOHO database [1]. The feature vector as a shorttime representation of speech consists of 16 LPCCepstral coefficients, as well as residual information appended in the form of a pitch estimate and a measure of vocality of the speech. The robustness in verification accuracy is also studied. The results show that while speech coders, G.729 in particular, introduce coding distortions that lead to verification performance degradation, proper augmented use of unconventional information nevertheless leads to a competitive performance on par with that of a wellstudied traditional system which does not involve signal coding and transmission. The result suggests that speaker verification over a cell phone connection remains feasible even though the signal has been encoded to 8 Kb/s.
more …
By
Bravo, Sergio; Calderón, Felix
Many of the watermarking schemes that claim resilience to geometrical distortions embed information into invariant or semiinvariant domains. However, the discretisation process required in such domains might lead to low correlation responses during watermarking detection. In this document, a new strategy is proposed to provide resilience to strong Rotation, Scaling and Translation (RST) distortions. The proposed detection process is based on a Genetic Algorithm (GA) that maximises the correlation coefficient between the originally embedded watermark and the input image. Comparisons between a previous scheme, based on LogPolar Mapping (LPM), and the present approach are reported. Results show that even a simple insertion process provides more robustness, as well as a lower image degradation.
more …
By
MarínHernández, Antonio; Devy, Michel; AviñaCervantes, Gabriel
Scene interpretation and feature tracking in natural environments are very complex perceptual functions. Complexity lies on several factors, for example: the lack of control on illumination conditions and the presence of different textures in the environment. This paper presents a realtime method to track roads in natural environments. The scene is previously characterized and classified in different regions by a combined ICA and color segmentation method (not described in this paper). This method is not so fast to track desired features in real time. The region tracking is executed on color active contours. New color potential fields are proposed: a) one to attract active contours depending on the selected region color, and b) the second one to repulse active contours when it is inside the region. Two potential fields are defined from the results of the initial characterization process and are updated by the same process at a given constant frequency, to avoid errors mainly due to global changes in illumination conditions or to local changes on the characteristics of the selected region. This approach has been evaluated on image sequences, acquired in natural environments.
more …
By
CruzRamírez, Nicandro; AcostaMesa, HéctorGabriel; BarrientosMartínez, RocíoErandi; NavaFernández, LuisAlonso
Show all (4)
In this paper, we evaluate the effectiveness of four Bayesian network classifiers as potential tools for the histopathological diagnosis of chronic idiopathic inflammatory bowel disease (CIIBD) using a database containing endoscopic colorectal biopsies. CIIBD is the generic term for referring to two ailments known as Crohn’s disease and ulcerative colitis. The results show that the defined histological attributes, considered relevant in the medical literature for the diagnosis of CIIBD, are very good for the distinction between normal samples and CIIBD samples (Crohn’s disease and ulcerative colitis combined into a single category) but less good for the explicit distinction between Crohn’s disease and ulcerative colitis. The findings suggest an intrinsic impossibility of selecting a set of features for achieving good balance for both sensitivity and specificity for Crohn’s disease and ulcerative colitis.
more …
By
HuertaHernández, Luis D.; ReyesGarcía, Carlos A.
In this work we propose an algorithm for continuous speech segmentation with text independency. In our approach we do not use feature vectors in order to detect phoneme boundaries, instead we only make use of the intensity measure. Obtaining with this a remarkable reduction in the amount of information needed and simplified rules on the processing. In the process only a preemphasis filter, and one strategy based on a distance measure with normalized fuzzy memberships over the signal patterns are used. In the preliminary results the method reaches up to 77.54% of correct segmentation with a 20 msec. accuracy and an over segmentation rate near to 0%. The algorithm implementation, the experiments, as well as some results are shown.
more …
By
HernándezReyes, Edith; CarrascoOchoa, J. A.; MartínezTrinidad, J. Fco.
Tin Kam Ho and Ester Bernardò Mansilla in 2004 proposed to use data complexity measures to determine the domain of competition of the classifiers. They applied different classifiers over a set of problems of two classes and determined the best classifier for each one. Then for each classifier they analyzed how the values of some pairs of complexity measures were, and based on this analysis they determine the domain of competition of the classifiers. In this work, we propose a new method for selecting the best classifier for a given problem, based in the complexity measures. Some experiments were made with different classifiers and the results are presented.
more …
By
Guzmán, Enrique; Pogrebnyak, Oleksiy; Fernández, Luis Sánchez; YáñezMárquez, Cornelio
Show all (4)
4 Citations
One of the most serious problems in vector quantization is the high computational complexity at the encoding phase. This paper presents a new fast search algorithm for vector quantization based on Extended Associative Memories (FSAEAM). In order to obtain the FSAEAM, first, we used the Extended Associative Memories (EAM) to create an EAMcodebook applying the EAM training stage to the codebook produced by the LBG algorithm. The result of this stage is an associative network whose goal is to establish a relation between training set and the codebook generated by the LBG algorithm. This associative network is EAMcodebook which is used by the FSAEAM. The FSAEAM VQ process is performed using the recalling stage of EAM. This process generates a set of the class indices to which each input vector belongs. With respect to the LBG algorithm, the main advantage offered by the proposed algorithm is high processing speed and low demand of resources (system memory), while the encoding quality remains competitive.
more …
By
Kober, Vitaly; Mozerov, Mikhail; ÁlvarezBorrego, Josué
1 Citations
An effective algorithm for automatic removal impulse noise from highly corrupted monochromatic images is proposed. The method consists of two steps. Outliers are first detected using local spatial relationships between image pixels. Then the detected noise pixels are replaced with the output of a rankorder filter over a local spatially connected area excluding the outliers, while noisefree pixels are left unaltered. Simulation results in test images show a superior performance of the proposed filtering algorithm comparing with conventional filters. The comparisons are made using mean square error, mean absolute error, and subjective human visual error criterion.
more …
By
RodríguezColín, Raúl; CarrascoOchoa, J. A.; MartínezTrinidad, J. Fco.
The KNN rule has been widely used in many pattern recognition problems, but it is sensible to noisy data within the training set, therefore, several sample edition methods have been developed in order to solve this problem. A. Franco, D. Maltoni and L. Nanni proposed the RewardPunishment Editing method in 2004 for editing numerical databases, but it has the problem that the selected prototypes could belong neither to the sample nor to the universe. In this work, we propose a modification based on selecting the prototypes from the training set. To do this selection, we propose the use of the Fuzzy Cmeans algorithm for mixed data and the KNN rule with similarity functions. Tests with different databases were made and the results were compared against the original RewardPunishment Editing and the whole set (without any edition).
more …
By
Guerra, Sergio Suárez; Rodríguez, José Luis Oropeza; Riveron, Edgardo M. Felipe; Nazuno, Jesús Figueroa
Show all (4)
1 Citations
This paper presents an approach for the automatic speech recognition using syllabic units. Its segmentation is based on using the ShortTerm Total Energy Function (STTEF) and the Energy Function of the High Frequency (ERO parameter) higher than 3,5 KHz of the speech signal. Training for the classification of the syllables is based on ten related Spanish language rules for syllable splitting. Recognition is based on a Continuous Density Hidden Markov Models and the bigram model language. The approach was tested using two voice corpus of natural speech, one constructed for researching in our laboratory (experimental) and the other one, the corpus Latino40 commonly used in speech researches. The use of ERO parameter increases speech recognition by 5% when compared with recognition using STTEF in discontinuous speech and improved more than 1.5% in continuous speech with three states. When the number of states is incremented to five, the recognition rate is improved proportionally to 97.5% for the discontinuous speech and to 80.5% for the continuous one.
more …
By
Calderon, Felix; Romero, Leonardo; Flores, Juan
2 Citations
We present the GA–SSD–ARC–NLM, a new robust parametric image registration technique based on the non–parametric image registration SSD–ARC algorithm. This new algorithm minimizes a new cost function quite different to the original nonparametric SSDARC, which explicitly models outlier punishments, using a combination of a genetic algorithm and the Newton–Levenberg–Marquardt method. The performance of the new method was compared against two robust registration techniques: the Lorentzian Estimator and the RANSAC method. Experimental tests using gray level images with outliers (noise) were done using the three algorithms. The goal was to find an affine transformation to match two images; the new method improves the other methods when noisy images are used.
more …
By
Vazquez, Roberto A.; Sossa, Humberto; Garro, Beatriz A.
1 Citations
Several associative memories (AM) have been proposed in the last years. These AMs have several constraints that limit their applicability in complex problems such as face recognition. Despite of the power of these models, they cannot reach its full power without applying new mechanisms based on current and future studies on biological neural networks. In this research we show how a network of dynamic associative memories (DAM) combined with some aspects of the infant vision system could be efficiently applied to the face recognition problem. Through several experiments by using a benchmark of faces the accuracy of the proposal is tested.
more …
By
FelipeRiveron, Edgardo; GarciaGuimeras, Noel
3 Citations
This paper presents a strategy for the extraction of blood vessels from ophthalmoscopic color images of the fundus of human retinas. To extract the vascular network, morphology operators were used, primarily maximum of openings and sum of valleys, and secondly a reconstruction by dilation from two images obtained using threshold by hysteresis. To extract the skeleton of the resulting vascular network, morphological thinning and pruning algorithms were used. Results obtained represent a starting point for future work related to the detection of anomalies in the vascular network and techniques for personal authentication.
more …
By
GarcíaPerera, L. Paola; NolazcoFlores, Juan A.; MexPerera, Carlos
1 Citations
In this research we propose to use phoneme spotting to improve the results in the generation of a cryptographic key. Phoneme spotting selects the phonemes with highest accuracy in the user classification task. The key bits are constructed by using the Automatic Speech Recognition and Support Vector Machines. Firstly, a speech recogniser detects the phoneme limits in each speech utterance. Afterwards, the support vector machine performs a user classification and generates a key. By selecting the highest accuracy phonemes for a a set of 10, 20, 30 and 50 speakers randomly chosen from the YOHO database, it is possible to generate reliable cryptographic keys.
more …
By
SánchezCruz, Hermilo
1 Citations
A new method for corner detection is proposed. Previous approaches for detecting corners rely on computing angle functions to find changes of curvature. Generally, those methods employ eight different symbols to represent contour shapes. The method of this work is based on using three symbols of a chain code to find pattern substrings, detecting corners in the contour shape. The method relies on searching for the relationship among neighbor points, finding two basic pattern contour chain elements, requiring few computing power to obtain shape corners.
more …
By
Sossa, Humberto; Barrón, Ricardo
1 Citations
Median associative memories (MEDMEMs) first described in [1] have proven to be efficient tools for the reconstruction of patterns corrupted with mixed noise. First formal conditions under which these tools are able to reconstruct patterns either from the fundamental set of patterns and from distorted versions of them were given in [1]. In this paper, new more accurate conditions are provided that assure perfect reconstruction. Numerical and real examples are also given.
more …
By
KuriMorales, Angel
5 Citations
In this paper we describe a heuristic approach to the problem of identifying a pattern embedded within a figure from a predefined set of patterns via the utilization of a genetic algorithm (GA). By applying this GA we are able to recognize a set of simple figures independently of scale, translation and rotation. We discuss the fact that this GA is, purportedly, the best among a set of alternatives; a fact which was previously proven appealing to statistical techniques. We describe the general process, the special type of genetic algorithm utilized, report some results obtained from a test set and we discuss the aforementioned results and we comment on these. We also point out some possible extensions and future directions.
more …
By
Sánchez, Pedro; Yáñez, Cornelio; Pecero, Jonathan; Ramírez, Apolinar
Show all (4)
In this paper a probabilistic approach is considered to develop a methodology to solve the problem of estimation of the position of the observer. The base of this methodology is the appearance vision with which an environment map is constructed using Kernel PCA. For the experiments an image set is acquired in unknown locations in the same environment. The performance of Kernel PCA technique was tested according to the optimum dimension of the environment model and the quantity of images correctly classified using a Bayesian algorithm. To validate the results obtained with Kernel PCA the same experiments were performed with PCA and APEX techniques, then the results were compared showing that Kernel PCA has better performance than PCA and APEX.
more …
By
López Gutiérrez, Luis David; Altamirano Robles, Leopoldo
In this paper an approach to the automatic target detection and tracking using multisensor image sequences with the presence of camera motion is presented. The approach consists of three parts. The first part uses a motion segmentation method for targets detection in the visible images sequence. The second part uses a background model for detecting objects presented in the infrared sequence, which is preprocessed to eliminate the camera motion. The third part combines the individual results of the detection systems; it extends the Joint Probabilistic Data Association (JPDA) algorithm to handle an arbitrary number of sensors. Our approach is tested using image sequences with high clutter on dynamic environments. Experimental results show that the system detects 99% of the targets in the scene, and the fusion module removes 90% of the false detections.
more …
By
HeroldGarcía, Silena; RiveraRovelo, Jorge; BayroCorrochano, Eduardo
1 Citations
One necessary task in the operating room is to establish a common reference frame, in order to relate the information obtained from different sensors, and to combine both the preoperative with the intraoperative information. To estimate the transformations between different data, fiducial markers are typically used. In this paper we present a formulation of the known handeye calibration problem, to estimate the transformation between an endoscopic camera and the set of spherical markers placed on it, using the conformal geometric algebra framework. Such markers are tracked by an optical stereo tracking system, which help to relate the real world with the virtual model created before surgery. Experimental results shows that our method is reliable and useful for medical applications in real time like neurosurgery.
more …
By
LópezEscobar, Saúl; CarrascoOchoa, J. A.; MartínezTrinidad, J. Fco.
1 Citations
The kmeans algorithm is a frequently used algorithm for solving clustering problems. This algorithm has the disadvantage that it depends on the initial conditions, for that reason, the global kmeans algorithm was proposed to solve this problem. On the other hand, the kmeans algorithm only works with numerical features. This problem is solved by the kmeans algorithm with similarity functions that allows working with qualitative and quantitative variables and missing data (mixed and incomplete data). However, this algorithm still depends on the initial conditions. Therefore, in this paper an algorithm to solve the dependency on initial conditions of the kmeans algorithm with similarity functions is proposed, our algorithm is tested and compared against kmeans algorithm with similarity functions.
more …
By
Fernandez, Luis Pastor Sanchez; Charles, Roberto Herrera; Pogrebnyak, Oleksiy
The wave is a complex and important phenomenon for structures de signs in the coastal zones and beaches. This paper presents a novel system for the generation of spectral patterns of unidirectional irregular waves in research laboratories. The system’s control basic elements are a linear motor, a servo controller and a personal computer. The used main mathematical tools are a feed forward neural network, digital signal processing and statistical analysis. The research aim is to obtain a system of more accuracy and small response time. This behavior is interpreted, in marine hydraulics, as a fast calibration of experiments. The wave power spectrums are generated in a test channel of rectangular section with dimensions: length 12 m; depth 40 cm; width 30 cm.
more …
By
Lira, Jorge
Abstract
A model of desertification in semiarid environment employing satellite multispectral images is presented. The variables proposed to characterize desertification are: texture of terrain, vegetation index for semiarid terrain, and albedo of terrain. The texture is derived from a divergence operator applied upon the vector field formed by the first three principal components of the image. The vegetation index selected is the TSAVI, suitable for semiarid environment where vegetation is scarce. The albedo is calculated from the first principal component obtained from the bands of the multispectral image. These three variables are input into a clustering algorithm resulting in six desertification grades. These grades are ordered from nodesertification to severe desertification. Details are provided for the computer calculation of the desertification variables, and the parameters employed in the clustering algorithm. A multispectral Landsat TM image is selected for this research. A thematic map of desertification is then generated with the support of ancillary data related to the study area.
more …
By
Rojo Ruiz, Arturo; Sánchez Fernandez, Luis P.; FelipeRiverón, Edgardo; Suárez Guerra, Sergio
Show all (4)
2 Citations
This paper presents a novel computational multimodal model designed for pattern recognition of aircrafts’ noise in real environments; with an 88.5% of effectiveness, it considers 13 different categories of aircrafts. This method includes measurements of signals of the noise produced during the takeoff at 25,000 samples per second and with a resolution of 24 bits, an spectral analysis made by means of an autoregressive model, an octave analysis, a normalization method created specifically for this work and two feedforward neural networks. All the signals used for the design and evaluation of the results were obtained by means of field measurements.
more …
By
KuriMorales, Angel
Abstract
In this paper we describe a heuristic approach to the problem of identifying a pattern embedded within a figure from a predefined set of patterns via the utilization of a genetic algorithm (GA). By applying this GA we are able to recognize a set of simple figures independently of scale, translation and rotation. We discuss the fact that this GA is, purportedly, the best among a set of alternatives; a fact which was previously proven appealing to statistical techniques. We describe the general process, the special type of genetic algorithm utilized, report some results obtained from a test set and we discuss the aforementioned results and we comment on these. We also point out some possible extensions and future directions.
more …
By
PonsPorrata, Aurora; Díaz, Guillermo Sánchez; Cortés, Manuel Lazo; Ramírez, Leydis Alfonso
Show all (4)
In this paper, we present an incremental clustering algorithm in the logical combinatorial approach to pattern recognition, which finds incrementally the β_{0}compact sets with radius α of an object collection. The proposed algorithm allows generating an intermediate subset of clusters between the β_{0}connected components and β_{0}compact sets (including both of them as particular cases). The evaluation experiments on standard document collections show that the proposed algorithm outperforms the algorithms that obtain the β_{0}connected components and the β_{0}compact sets.
more …
