Skip to main content
main-content

Über dieses Buch

The two-volume set LNCS 5601 and LNCS 5602 constitutes the refereed proceedings of the Third International Work-Conference on the Interplay between Natural and Artificial Computation, IWINAC 2009, held in Santiago de Compostela, Spain, in June 2009. The 108 revised papers presented are thematically divided into two volumes. The first volume includes papers relating the most recent collaborations with Professor Mira and contributions mainly related with theoretical, conceptual and methodological aspects linking AI and knowledge engineering with neurophysiology, clinics and cognition. The second volume contains all the contributions connected with biologically inspired methods and techniques for solving AI and knowledge engineering problems in different application domains.

Inhaltsverzeichnis

Frontmatter

Measurements over the Aquiles Tendon through Ecographic Images Processing

Boundary detection has a relevant importance in locomotor system ecographies, mainly because some illnesses and injuries can be detected before the first symptoms appear. The images used show a great variety of textures as well as non clear edges. This drawback may result in different contours depending on the person who traces them out and different diagnoses too. This paper presents the results of applying the geodesic active contour and other boundary detection techniques in ecographic images of Aquiles tendon, such as morphological image processing and active contours. Other modifications to this algorithm are introduced, like matched filtering. In order to upgrade the smoothness of the final contour, morphological image processing and polynomial interpolation has been used with great results. Actually, the automatization of boundary detection improves the measurement procedure, obtaining error rates under ±10%.

M-Consuelo Bastida-Jumilla, Juan Morales-Sánchez, Rafael Verdú-Monedero, Jorge Larrey-Ruiz, José Luis Sancho-Gómez

A New Approach in Metal Artifact Reduction for CT 3D Reconstruction

The 3D representation of CT scans is widely used in medical application such as virtual endoscopy, plastic reconstructive surgery, dental implant planning systems and more. Metallic objects present in CT studies cause strong artifacts like beam hardening and streaking, what difficult to a large extent the 3D reconstruction. Previous works in this field use projection data in different ways with the aim of artifact reduction. But in DICOM-based applications this information is not available, thus the need for a new point of view regarding this issue. Our aim is to present an exhaustive study of the state of the art and to evaluate a new approach based in mathematical morphology in polar domain in order to reduce the noise but preserving dental structures, valid for real-time applications.

Valery Naranjo, Roberto Llorens, Patricia Paniagua, Mariano Alcañiz, Salvador Albalat

Genetic Approaches for the Automatic Division of Topological Active Volumes

The Topological Active Volumes is an active model focused on 3D segmentation tasks. It is based on the 2D Topological Active Nets model and provides information about the surfaces and the inside of the detected objects in the scene. This paper proposes new optimization approaches based on Genetic Algorithms combined with a greedy local search that improve the results of the 3D segmentations and overcome some drawbacks of the model related to parameter tuning or noise conditions. The hybridization of the genetic algorithm with the local search allows the treatment of topological changes in the model, with the possibility of an automatic subdivision of the Topological Active Volume. This combination integrates the advantages of the global and local search procedures in the segmentation process.

J. Novo, N. Barreira, M. G. Penedo, J. Santos

Object Discrimination by Infrared Image Processing

Signal processing applied to pixel by pixel infrared image processing has been frequently used as a tool for fire detection in different scenarios. However, when processing the images pixel by pixel, the geometrical or spatial characteristics of the objects under test are not considered, thus increasing the probability of false alarms. In this paper we use classical techniques of image processing in the characterization of objects in infrared images. While applying image processing to thermal images it is possible to detect groups of hotspots representing possible objects of interest and extract the most suitable features to distinguish between them. Several parameters to characterize objects geometrically, such as fires, cars or people, have been considered and it has been shown their utility to reduce the probability of false alarms of the pixel by pixel signal processing techniques.

Ignacio Bosch, Soledad Gomez, Raquel Molina, Ramón Miralles

Validation of Fuzzy Connectedness Segmentation for Jaw Tissues

Most of the dental implant planning systems implement 3D reconstructions of the CT-data in order to achieve more intuitive interfaces. This way, the dentists or surgeons can handle the patient’s virtual jaw in the space and plan the location, orientation and some other features of the implant from the orography and density of the jaw. The segmentation of the jaw tissues (the cortical bone, the trabecular core and the mandibular channel) is critical for this process, because each one has different properties and in addition, because an injury of the channel in the surgery may cause lip numbness. Current programs don’t carry out the segmentation process or just do it by hard thresholding or by means of exhaustive human interaction. This paper deals with the validation of fuzzy connectedness theory for the automated, accurate and time efficient segmentation of jaw tissues.

Roberto Lloréns, Valery Naranjo, Miriam Clemente, Mariano Alcañiz, Salvador Albalat

Breast Cancer Classification Applying Artificial Metaplasticity

In this paper we are apply Artificial Metaplasticity MLP (MMLPs) to Breast Cancer Classification. Artificial Metaplasticity is a novel ANN training algorithm that gives more relevance to less frequent training patterns and subtract relevance to the frequent ones during training phase, achieving a much more efficient training, while at least maintaining the Multilayer Perceptron performance. Wisconsin Breast Cancer Database (WBCD) was used to train and test MMLPs. WBCD is a well-used database in machine learning, neural networks and signal processing. Experimental results show that MMLPs reach better accuracy than any other recent results.

Alexis Marcano-Cedeño, Fulgencio S. Buendía-Buendía, Diego Andina

Ontology Based Approach to the Detection of Domestics Problems for Independent Senior People

In the first decade of the 21st century, there is a tremendous increment in the number of elderly people which live independently in their own houses. In this work, we focus on elderly people which spend almost all the time by their own. The goal of this work is to build an artificial system capable of unobtrusively monitor this concrete subject. In this case, the system must be capable of detecting potential situations of danger (e.g. the person lays unmobilised in the floor or she is suffering some kind of health crysis). This is done without any wearable device but only using a sensor network and an intelligent processing unit within a single and small CPU. This kind of such unbostrusive system makes seniors to augment his or her perception of independence and safeness at home.

Juan A. Botia Blaya, Jose Palma, Ana Villa, David Perez, Emilio Iborra

A Wireless Sensor Network for Assisted Living at Home of Elderly People

This paper introduces an ubiquitous wireless network infrastructure to support an assisted living at home system. This system integrates a set of smart sensors which are designed to provide care assistence and security to elderly citizens living at home alone. The system facilitates privacy by performing local computation, it supports heterogeneous sensor devices and it provides a platform and initial architecture for exploring the use of sensors with elderly people. We have developed a low-power multihop network protocol consists of nodes (Motes) that wirelessly communicate to each other and are capable of hopping radio messages to a base station where they are passed to a PC (or other possible client). The goal of this project is to provide alerts to caregivers in the event of an accident, acute illness or strange (possibly dangerous) activities, and enable monitoring by authorized and authenticated caregivers. In this paper, we describe ubiquitous assistential monitoring system at home. We have focused on the unobtrusive habitual activities signal measurement and wireless data transfer using ZigBee technology.

Francisco Fernández-Luque, Juan Zapata, Ramón Ruiz, Emilio Iborra

An Ambient Assisted Living System for Telemedicine with Detection of Symptoms

Elderly people have a high risk of health problems. Hence, we propose an architecture for Ambient Assisted Living (AAL) that supports pre-hospital health emergencies, remote monitoring of patients with chronic conditions and medical collaboration through sharing of health-related information resources (using the European electronic health records CEN/ISO EN13606). Furthermore, it is going to use medical data from vital signs for, on the one hand, the detection of symptoms using a simple rule system (e.g. fever), and on the other hand, the prediction of illness using chronobiology algorithms (e.g. prediction of myocardial infarction eight days before). So this architecture provides a great variety of communication interfaces to get vital signs of patients from a heterogeneous set of sources, as well as it supports the more important technologies for Home Automation. Therefore, we can combine security, comfort and ambient intelligence with a telemedicine solution, thereby, improving the quality of life in elderly people.

A. J. Jara, M. A. Zamora-Izquierdo, A. F. Gomez-Skarmeta

Applying Context-Aware Computing in Dependent Environments

Context-aware systems gather data from their surrounding environments in order to offer completely new opportunities in the development of end user applications. Used in conjunction with mobile devices, these systems are of great value and increase usability. Applications and services should adapt to changing conditions within dynamic environments. This article analyzes the important aspects of context-aware computing and shows how it can be applied to monitor dependent individuals in their home. The proposed system logically processes the data it receives in order to identify and maintain a permanent location on the patient in the home, managing the infrastructure of services both safely and securely.

Juan A. Fraile, Javier Bajo, Juan M. Corchado

A Smart Solution for Elders in Ambient Assisted Living

Ambient Assisted Living (AAL) includes assistance to carry out daily activities, health and activity monitoring, enhancing safety and security, getting access to, medical and emergency systems. Ambient home care systems (AHCS) are specially design for this purpose; they aim at minimizing the potential risks that living alone may suppose for an elder, thanks to their capability of gathering data of the user, inferring information about his activity and state, and taking decisions on it. In this paper, we present several categories of context-aware services. One related to the autonomy enhancement including services like: medication, shopping and cooking. And another which is the emergency assistant category designed for the assistance, prediction and prevention of any emergency occurred addressed to any elder and their caregivers. These services run on the top of an AHCS, which collects data from a network of environmental, health and physical sensors and then there is a context engine, customized on Appear platform that holds the inference and reasoning functionalities.

Nayat Sánchez-Pi, José Manuel Molina

Convergence of Emergent Technologies for the Digital Home

The Digital Home is the result of the convergence of technologies of different nature that interact with each other in the Home environment. It is a realization of the Ambient Intelligence concept. The final objective of the Ambient Intelligence is that sensors, devices and networks that compose this environment can co-exist with human users, to improve their quality of live. The relevant characteristic of the Digital Home as a main scenario of Ambient Intelligence is its pervasive nature. This paper describes these technologies and their harmonization, based on the work done in the INREDIS project, which deals with accessibility and new technologies.

Celia Gutiérrez, Sara Pérez

Results of an Adaboost Approach on Alzheimer’s Disease Detection on MRI

In this paper we explore the use of the Voxel-based Morphometry (VBM) detection clusters to guide the feature extraction processes for the detection of Alzheimer’s disease on brain Magnetic Resonance Imaging (MRI). The voxel location detection clusters given by the VBM were applied to select the voxel values upon which the classification features were computed. We have evaluated feature vectors computed over the data from the original MRI volumes and from the GM segmentation volumes, using the VBM clusters as voxel selection masks. We use the Support Vector Machine (SVM) algorithm to perform classification of patients with mild Alzheimer’s disease vs. control subjects. We have also considered combinations of isolated cluster based classifiers and an Adaboost strategy applied to the SVM built on the feature vectors. The study has been performed on MRI volumes of 98 females, after careful demographic selection from the Open Access Series of Imaging Studies (OASIS) database, which is a large number of subjects compared to current reported studies. Results are moderately encouraging, as we can obtain up to 85% accuracy with the Adaboost strategy in a 10-fold cross-validation.

Alexandre Savio, Maite García-Sebastián, Manuel Graña, Jorge Villanúa

Analysis of Brain SPECT Images for the Diagnosis of Alzheimer Disease Using First and Second Order Moments

This paper presents a computer-aided diagnosis technique for improving the accuracy of the early diagnosis of the Alzheimer type dementia. The proposed methodology is based on the selection of the voxels which present greater overall difference between both modalities (normal and Alzheimer) and also lower dispersion. We measure the dispersion of the intensity values for normals and Alzheimer images by mean of the standard deviation images. The mean value of the intensities of selected voxels is used as feature for different classifiers, including support vector machines with linear kernels, fitting a multivariate normal density to each group and the k-nearest neighbors algorithm. The proposed methodology reaches an accuracy of 92 % in the classification task.

D. Salas-Gonzalez, J. M. Górriz, J. Ramírez, M. López, I. Álvarez, F. Segovia, C. G. Puntonet

Neurobiological Significance of Automatic Segmentation: Application to the Early Diagnosis of Alzheimer’s Disease

Alzheimer’s disease is a progressive neurodegenerative disease that affects particularly memory function. Specifically, the neural system responsible for encoding and retrieval of the memory for facts and events (declarative memory) is dependent on anatomical structures located in the medial part of the temporal lobe (MTL). Clinical lesions as well as experimental evidence point that the hippocampal formation (hippocampus plus entorhinal cortex) and the adjacent cortex, both main components of the MTL, are the regions critical for normal declarative memory function. Neuroimage studies as ours, have taken advantage of the feasibility of manual segmentation of the gray matter volume, which correlates with memory impairment and clinical deterioration of Alzheimer’s disease patients. We wanted to explore the advantages of automatic segmentation tools, and present results based on one 3T MRI in a young subject. The automatic segmentation allowed a better discrimination between extracerebral structures and the surface of the brain, as well as an improvement both in terms of speed and reliability in the demarcation of different MTL structures, all of which play a key role in declarative memory processing. Based largely on our own nonhuman primate data on brain and hippocampal connections, we defined automatically the angular bundle in the MTL as the fibers containing the perforant path (interconnection and dialogue between the entorhinal cortex and its hippocampal termination. The speed and accuracy of the technique needs further development, but it seems to be promising enough for early detection of memory deficits associated to Alzheimer’s disease.

Ricardo Insausti, Mariano Rincón, César González-Moreno, Emilio Artacho-Pérula, Amparo Díez-Peña, Tomás García-Saiz

Support Vector Machines and Neural Networks for the Alzheimer’s Disease Diagnosis Using PCA

In the Alzheimer’s Disease (AD) diagnosis process, functional brain images such as Single-Photon Emission Computed Tomography (SPECT) and Positron Emission Tomography (PET) have been widely used to guide the clinicians. However, the current evaluation of these images entails a succession of manual reorientations and visual interpretation steps, which attach in some way subjectivity to the diagnostic. In this work, two pattern recognition methods have been applied to SPECT and PET images in order to obtain an objective classifier which is able to determine whether the patient suffers from AD or not. A common feature selection stage is first described, where Principal Component Analysis (PCA) is applied over the data to drastically reduce the dimension of the feature space, followed by the study of neural networks and support vector machines (SVM) classifiers. The achieved accuracy results reach 98.33% and 93.41% for PET and SPECT respectively, which means a significant improvement over the results obtained by the classical Voxels-As-Features (VAF) reference approach.

M. López, J. Ramírez, J. M. Górriz, I. Álvarez, D. Salas-Gonzalez, F. Segovia, M. Gómez-Río

Functional Brain Image Classification Techniques for Early Alzheimer Disease Diagnosis

Currently, the accurate diagnosis of the Alzheimer disease (AD) still remains a challenge in the clinical practice. As the number of AD patients has increased, its early diagnosis has received more attention for both social and medical reasons. Single photon emission computed tomography (SPECT), measuring the regional cerebral blood flow, enables the diagnosis even before anatomic alterations can be observed by other imaging techniques. However, conventional evaluation of SPECT images often relies on manual reorientation, visual reading and semiquantitative analysis of certain regions of the brain. This paper evaluates different pattern classifiers including

k

-nearest neighbor (

k

NN), classification trees, support vector machines and feedforward neural networks in combination with template-based normalized mean square error (NMSE) features of several coronal slices of interest (SOI) for the development of a computer aided diagnosis (CAD) system for improving the early detection of the AD. The proposed system, yielding a 98.7% AD diagnosis accuracy, reports clear improvements over existing techniques such as the voxel-as-features (VAF) which yields just a 78% classification accuracy.

J. Ramírez, R. Chaves, J. M. Górriz, I. Álvarez, M. López, D. Salas-Gonzalez, F. Segovia

Quality Checking of Medical Guidelines Using Interval Temporal Logics: A Case-Study

Computer-based decision support in health-care is becoming more and more important in recent years.

Clinical Practise Guidelines

are documents supporting health-care professionals in managing a disease in a patient, in order to avoid non-standard practices or outcomes. In this paper, we consider the problem of formalizing a guideline in a logical language. The target language is an interval-based temporal logic interpreted over natural numbers, namely the Propositional Neighborhood Logic, which has been shown to be expressive enough for our objective, and for which the satisfiability problem has been shown to be decidable. A case-study of a real guideline is presented.

Guido Sciavicco, Jose M. Juarez, Manuel Campos

Classification of SPECT Images Using Clustering Techniques Revisited

We present a novel classification method of SPECT images based on clustering for the diagnosis of Alzheimer’s disease. The aims of the clustering approach which is based on Gaussian Mixture Model (GMM) for density estimation, is to automatically select Regions of Interest (ROIs) and to effectively reduce the dimensionality of the problem. The clusters represented by Gaussians are constructed according to a maximum likelihood criterion employing the expectation maximization (EM) algorithm. By considering only the intensity levels inside the clusters, the resulting feature space has a significantly reduced dimensionality with respect to former approaches using the voxel intensities directly as features. With this feature extraction method one avoids the so-called small sample size problem and nonlinear classifiers may be used to distinguish between the brain images of normal and Alzheimer patients. Our results show that for various classifiers the clustering method yields higher accuracy rates than the classification considering all voxel values.

J. M. Górriz, J. Ramírez, A. Lassl, I. Álvarez, F. Segovia, D. Salas, M. López

Detection of Microcalcifications Using Coordinate Logic Filters and Artificial Neural Networks

Breast cancer is one of the leading causes to women mortality in the world. Cluster of Microcalcifications (MCC) in mammograms can be an important early sign of breast cancer, the detection is important to prevent and treat the disease. In this paper, we present a novel method for the detection of MCC in mammograms which consists of image enhancement by histogram adaptive equalization technique, MCC edge detection by Coordinate Logic Filters (CLF), generation, clustering and labelling of suboptimal features vectors by means of Self Organizing Map (SOM) Neural Network. Like comparison we applied an unsupervised clustering K-means in the stage of labelling of our method. In the labelling stage, we obtain better results with the proposed SOM Neural Network compared with the k-means algorithm. Then, we show that the proposed method can locate MCCs in an efficient way.

J. Quintanilla-Domínguez, M. G. Cortina-Januchs, J. M. Barrón-Adame, A. Vega-Corona, F. S. Buendía-Buendía, D. Andina

Rule Evolving System for Knee Lesion Prognosis from Medical Isokinetic Curves

This paper proposes a system for applying data mining to a set of time series with medical information. The series represent an isokinetic curve that is obtained from a group of patients performing a knee exercise on an isokinetic machine. This system has two steps: the first one is to analyze the input time series in order to generate a simplified model of an isokinetic curve; the second step applies a grammar-guided genetic program including an evolutionary gradient operator and an entropy-based fitness function to obtain a set of rules for a knowledge-based system. This system performs medical prognosis for knee injury detection. The results achieved have been statistically compared to another evolutionary approach that generates fuzzy rule-based systems.

Jorge Couchet, José María Font, Daniel Manrique

Denoising of Radiotherapy Portal Images Using Wavelets

Verifying that each radiation beam is being delivered as intended constitutes a fundamental issue in radiation therapy. In order to verify the patient positioning the own high energy radiation beam is commonly used to produce an image similar to a radiography (portal image), which is further compared to an image generated in the simulation-planning phase. The evolution of radiotherapy is parallel to the increase of the number of portal images used for each treatment, and as a consecuence the radiation dose due to image has increased notably. The concern arises from the fact that the radiation delivered during imaging is not confined to the treatment volumes. One possible solution should be the reduction of dose per image, but the image quality should become lower as the quantum noise should become higher. The limited quality of portal images makes difficult to propose dose reduction if there is no way to deal with noise increment. In this work we study the denoising of portal images by various denoising algorithms. In particular we are interested in wavelet-based denoising. The wavelet-based algorithms used are the shrinkage by wavelet coefficients thresholding, the coefficient extraction based on correlation between wavelet scales and the Bayesian least squares estimate of wavelet coefficients with Gaussians scale mixtures as priors (BLS-GSM). Two algorithms that do not use wavelets are also evaluated, a local Wiener estimator and the Non Local Means algorithm (NLM). We found that wavelet thresholding, wavelet coefficients extraction after correlation and NLM reach higher values of ISNR than Local Wiener. Also, the highest ISNR is reached by the BLS-GSM algorithm. This algorithm also produces the best visual results. We believe that these results are very encouraging for exploring forms of reducing the radiation doses associated to portal image in radiotherapy.

Antonio González-López, Juan Morales-Sánchez, María-Consuelo Bastida-Jumilla, Francisco López-Sánchez, Bonifacio Tobarra-González

A Block-Based Human Model for Visual Surveillance

This paper presents BB6-HM, a block-based human model for real-time monitoring of a large number of visual events and states related to human activity analysis, which can be used as components of a library to describe more complex activities in such important areas as surveillance. BB6-HM is inspired by the proportionality rules commonly used in Visual Arts, i.e., for dividing the human silhouette into six rectangles of the same height. The major advantage of this proposal is that analysis of the human can be easily broken down into parts, which allows us to introduce more specific domain knowledge and to reduce the computational load. It embraces both frontal and lateral views, is a fast and scale-invariant method and a large amount of task-focused information can be extracted from it.

Encarnación Folgado, Mariano Rincón, Margarita Bachiller, Enrique J. Carmona

Image Equilibrium: A Global Image Property for Human-Centered Image Analysis

Photographs and pictures created by humans present schemes and structures in their composition which can be analysed on semantic levels, irrespective of subject or content. The search for equilibrium in composition is a constant which enables us to establish a kind of image syntax, creating a visual alphabet from basic elements such as point, line, contour, texture, etc. This paper describes an operator which quantifies image equilibrium, providing a picture characterisation very close to a pixel matrix with considerable semantic content.

Ó. Sánchez, M. Rincón

Vision-Based Text Segmentation System for Generic Display Units

The increasing use of display units in avionics motivate the need for vision-based text recognition systems to assist humans. The system for generic displays proposed in this paper includes some of the usual text recognition steps, namely localization, extraction and enhancement, and optical character recognition. The proposal has been fully developed and tested on a multi-display simulator. The commercial OCR module from Matrox Imaging Library has been used to validate the textual displays segmentation proposal.

José Carlos Castillo, María T. López, Antonio Fernández-Caballero

Blind Navigation along a Sinuous Path by Means of the See ColOr Interface

The See ColOr interface transforms a small portion of a coloured video image into sound sources represented by spatialized musical instruments. This interface aims at providing visually impaired people with a capability of perception of the environment. In this work, the purpose is to verify the hypothesis that it is possible to use sounds from musical instruments to replace colour. Compared to state of the art devices, a quality of the See ColOr interface is that it allows the user to receive a feed-back auditory signal from the environment and its colours, promptly. An experiment based on a head mounted camera has been performed. Specifically, this experiment is related to outdoor navigation for which the purpose is to follow a sinuous path. Our participants successfully went along a red serpentine path for more than 80 meters.

Guido Bologna, Benoît Deville, Thierry Pun

Using Reconfigurable Supercomputers and C-to-Hardware Synthesis for CNN Emulation

The complexity of hardware design methodologies represents a significant difficulty for non hardware focused scientists working on CNN-based applications. An emerging generation of Electronic System Level (ESL) design tools is been developed, which allow software-hardware codesign and partitioning of complex algorithms from High Level Language (HLL) descriptions. These tools, together with High Performance Reconfigurable Computer (HPRC) systems consisting of standard microprocessors coupled with application specific FPGA chips, provide a new approach for rapid emulation and acceleration of CNN-based applications. In this article CoDeveloper, and ESL IDE from Impulse Accelerated Technologies, is analyzed. A sequential CNN architecture, suitable for FPGA implementation, proposed by the authors in a previous paper, is implemented using CoDeveloper tools and the DS1002 HPRC platform from DRC Computers. Results for a typical edge detection algorithm shown that, with a minimum development time, a 10x acceleration, when compared to the software emulation, can be obtained.

J. Javier Martínez-Álvarez, F. Javier Garrigós-Guerrero, F. Javier Toledo-Moreo, J. Manuel Ferrández-Vicente

Access Control to Security Areas Based on Facial Classification

The methods of biometric access control are currently booming due to increased security checks at business and organizational areas. Belong to this area applications based on fingerprints and iris of the eye, among others. However, although there are many papers related to facial recognition, in fact it is difficult to apply to real-world applications because of variations in lighting, position and changing expressions and appearance. In addition, systems proposed in the laboratory do not usually contain a large volume of samples, or the test variations not may be used in applications in real environments. Works include the issue of recognition of the individual, but not the access control based only on facial detect, although there are applications that combine cards with facial recognition, working more on the verification that identification. This paper proposes a robust system of classification based on a multilayer neural network, whose input will be samples of facial photographs with different variations of lighting, position and even time, with a volume of samples that simulates a real environment. Output is not the recognition of the individual, but the class to which it belongs. Through the experiments, it is demonstrated that this relatively simple structure is enough to select the main characteristics of the individuals, and, in the same process, enable the network to correctly classify individuals before entering the restricted area.

Aitor Moreno Fdz. de Leceta, Mariano Rincón

Comparing Feature Point Tracking with Dense Flow Tracking for Facial Expression Recognition

This work describes a research which compares the facial expression recognition results of two point-based tracking approaches along the sequence of frames describing a facial expression: feature point tracking and holistic face dense flow tracking. Experiments were carried out using the Cohn-Kanade database for the six types of prototypic facial expressions under two different spatial resolutions of the frames (the original one and the images reduced to a 40% of its original size). Our experimental results showed that the dense flow tracking method provided in average for the considered types of expressions a better recognition rate (95.45% of success) than feature point flow tracking (91.41%) for the whole test set of facial expression sequences.

José V. Ruiz, Belén Moreno, Juan José Pantrigo, Ángel Sánchez

A Memory-Based Particle Filter for Visual Tracking through Occlusions

Visual detection and target tracking are interdisciplinary tasks oriented to estimate the state of moving objects in an image sequence. There are different techniques focused on this problem. It is worth highlighting particle filters and Kalman filters as two of the most important tracking algorithms in the literature. In this paper, we presented a visual tracking algorithm which combines the particle filter framework with memory strategies to handle occlusions, called as memory-based particle filter (MbPF). The proposed algorithm follows the classical particle filter stages when a confidence measurement can be obtained from the system. Otherwise, a memory-based module try to estimate the hidden target state and to predict its future states using the process history. Experimental results showed that the performance of the MbPF is better than a standard particle filter when dealing with occlusion situations.

Antonio S. Montemayor, Juan José Pantrigo, Javier Hernández

Classification of Welding Defects in Radiographic Images Using an ANN with Modified Performance Function

In this paper, we describe an automatic classification system of welding defects in radiographic images. In a first stage, image processing techniques, including noise reduction, contrast enhancement, thresholding and labelling, were implemented to help in the recognition of weld regions and the detection of weld defects. In a second stage, a set of geometrical features which characterise the defect shape and orientation was proposed and extracted between defect candidates. In a third stage, an artificial neural network for weld defect classification was used under a regularisation process with different architectures for the input layer and the hidden layer. Our aim is to analyse this ANN modifying the performance function for differents neurons in the input and hidden layer in order to obtain a better performance on the classification stage.

Rafael Vilar, Juan Zapata, Ramón Ruiz

Texture Classification of the Entire Brodatz Database through an Orientational-Invariant Neural Architecture

This paper presents a supervised neural architecture, called SOON, for texture classification. Multi-scale Gabor filtering is used to extract the textural features which shape the input to a neural classifier with orientation invariance properties in order to accomplish the classification. Three increasing complexity tests over the well-known Brodatz database are performed to quantify its behavior. The test simulations, including the entire texture album classification, show the stability and robustness of the SOON response.

F. J. Díaz-Pernas, M. Antón-Rodríguez, J. F. Díez-Higuera, M. Martínez-Zarzuela, D. González-Ortega, D. Boto-Giralda

Eye-Hand Coordination for Reaching in Dorsal Stream Area V6A: Computational Lessons

Data related to the coordination and modulation between visual information, gaze direction and arm reaching movements in primates are analyzed from a computational point of view. The goal of the analysis is to construct a model of the mechanisms that allow humans and other primates to build dynamical representations of their peripersonal space through active interaction with nearby objects. The application of the model to robotic systems will allow artificial agents to improve their skills in their exploration of the nearby space.

Eris Chinellato, Beata J. Grzyb, Nicoletta Marzocchi, Annalisa Bosco, Patrizia Fattori, Angel P. del Pobil

Toward an Integrated Visuomotor Representation of the Peripersonal Space

The purpose of this work is the creation of a description of objects in the peripersonal space of a subject that includes two kinds of concepts, related to on-line, action-related features and memorized, conceptual ones, respectively. The inspiration of such description comes from the distinction between sensorimotor and perceptual visual processing as performed by the two visual pathways of the primate cortex. A model of such distinction, and of a further subdivision of the dorsal stream, is advanced with the purpose of applying it to a robotic setup. The model constitutes the computational basis for a robotic system able to achieve advanced skills in the interaction with its peripersonal space.

Eris Chinellato, Beata J. Grzyb, Patrizia Fattori, Angel P. del Pobil

Evidence for Peak-Shaped Gaze Fields in Area V6A: Implications for Sensorimotor Transformations in Reaching Tasks

The area V6A of the medial parieto-occipital cortex of the macaque is studied for gaze sensitivity. The reported experimental observations support the computational theory of the gain fields to produce a distributed representation of the real position of targets in head-centered coordinates. Although it was originally pointed out that the majority of the cells exhibit roughly linear gain fields [1] [2], we have verified that the peak-shaped gaze fields reported in this study are not in contrast with the gain field models developed in the theoretical neuroscience literature [3] [4]. Rather, the use of peak-shaped (e.g., non monotonic) gaze fields even improves the efficiency of the coding scheme by reducing the number of units that are necessary to encode the target position.

Rossella Breveglieri, Annalisa Bosco, Andrea Canessa, Patrizia Fattori, Silvio P. Sabatini

Segmenting Humans from Mobile Thermal Infrared Imagery

Perceiving the environment is crucial in any application related to mobile robotics research. In this paper, a new approach to real-time human detection through processing video captured by a thermal infrared camera mounted on the indoor autonomous mobile platform mSecurit

TM

is introduced. The approach starts with a phase of static analysis for the detection of human candidates through some classical image processing techniques such as image normalization and thresholding. Then, the proposal uses Lukas and Kanade optical flow without pyramids algorithm for filtering moving foreground objects from moving scene background. The results of both phases are compared to enhance the human segmentation by infrared camera. Indeed, optical flow will emphasize the foreground moving areas gotten at the initial human candidates detection.

José Carlos Castillo, Juan Serrano-Cuerda, Antonio Fernández-Caballero, María T. López

My Sparring Partner Is a Humanoid Robot

A Parallel Framework for Improving Social Skills by Imitation

This paper presents a framework for parallel tracking of human hands and faces in real time, and is a partial solution to a larger project on human-robot interaction which aims at training autistic children using a humanoid robot in a realistic non-restricted environment. In addition to the framework, the results of tracking different hand waving patterns are shown. These patterns provide an easy to understand profile of hand waving, and can serve as the input for a classification algorithm.

Tino Lourens, Emilia Barakova

Brain-Robot Interface for Controlling a Remote Robot Arm

This paper describes a technique based on electroencephalography (EEG) to control a robot arm. This technology could eventually allow people with severe disabilities to control robots that can help them in daily living activities. The EEG-based Brain Computer Interface (BCI) developed consists in register the brain rhythmic activity through a electrodes situated on the scalp in order to differentiate one cognitive process from rest state and use it to control one degree of freedom of the robot arm. In the paper the processing and classifier algorithm are described and an analysis of their parameters has been made with the objective of find the optimum configuration that allow obtaining the best results.

Eduardo Iáñez, M. Clara Furió, José M. Azorín, José Alejandro Huizzi, Eduardo Fernández

Learning to Coordinate Multi-robot Competitive Systems by Stimuli Adaptation

The area of competitive robotic systems usually yields to highly complicated strategies that must be achieved by complex learning architectures since analytic solutions seems to be unpractical or unfeasible at all. In this work we design an experiment in order to study and validate a model in the task of learning to coordinate a robot team to achieve complex goals by means of a simulation of a multi-robot competitive task that imitates a complex prey/predator system composed by three robots: predator, defender and prey. By means of such simulation we validate a general model about the complex phenomena of adaptation, anticipation and rationality.

José Antonio Martín H., Javier de Lope, Darío Maravall

A Behavior Based Architecture with Auction-Based Task Assignment for Multi-robot Industrial Applications

The study of collective robotic systems and how the interaction of the units that make them up can be harnessed to perform useful tasks is one of the main research topics in autonomous robotics. Inspiration for solutions in this realm can be sought in nature and in the interaction of natural social systems whether through simple trading strategies or through more complex economic models. Here we present a three level behavior based architecture for the implementation of multi-robot based cooperation systems that is based on the individual, the collective and the social levels. In particular, here we are going to consider the application of this architecture for the implementation and study of auction-based strategies for assigning tasks in a real application of multi-robot systems. Our approach is more focused on studying the behavior of auction-based techniques from an engineering point of view in terms of parameters and results analysis. To this end, we have used a real industrial case as an experimental platform where a heterogeneous group of robots must clean a ship tank. The results obtained show how the performance of the auction mechanism we have implemented does not degrade in terms of computational cost when the number of robots is increased, and how the complexity of the task assignment can be highly increased without any change in the cooperative control system.

Paula Garcia, Pilar Caamaño, Francisco Bellas, Richard J. Duro

On the Control of a Multi-robot System for the Manipulation of an Elastic Hose

The aim of this paper is to derive control strategies for a multi-robot system trying to move a flexible hose. We follow the approach of Geometric Exact Dynamic Splines to model the hose and its dynamics. The control problem is then stated as the problem of reaching a desired configuration of the spline control points from an initial configuration. The control of the hose by the multi-robot system is first solved neglecting the hose internal dynamics. We can derive the motion of the robot attachments that move that splines towards the desired configuration. Taking into account the dynamical model, we can derive the dynamic relations between the robots in the system and the motion of the hose towards the desired configuration.

Zelmar Echegoyen, Alicia d’Anjou, Manuel Graña

An Improved Evolutionary Approach for Egomotion Estimation with a 3D TOF Camera

We propose an evolutionary approach for egomotion estimation with a 3D TOF camera. It is composed of two main modules plus a preprocessing step. The first module computes the Neural Gas (NG) approximation of the preprocessed camera 3D data. The second module is an Evolution Strategy which performs the task of estimating the motion parameters by searching on the space of linear transformations restricted to the translation and rotation, applied on the codevector sets obtained by the NG for successive camera readings. The fitness function is the matching error between the transformed last set of codevectors and the codevector set corresponding to the next camera readings. In this paper, we report new modifications and improvements of this system and provide several comparisons between our and other well known registration algorithms.

Ivan Villaverde, Manuel Graña

A Frame for an Urban Traffic Control Architecture

Due to its potential for going into details or getting a global view of the system, agent architecture is a good frame to create an urban traffic control system. In fact, the agent architecture has allowed us to design a control system able of coordinating the traffic of a set of cars in certain scenarios, using, as initial core, the car control algorithms. In further steps, a higher level layer with the decision making systems and a lower level layer with the car control actuators have been added to the agents. Finally, the agent architecture can be extended with a higher level layers to control the traffic in critical areas or urban areas.

Teresa de Pedro, Ricardo García, Carlos González, Javier Alonso, Enrique Onieva, Vicente Milanés, Joshué Pérez

Partial Center of Area Method Used for Reactive Autonomous Robot Navigation

A new method for reactive autonomous robot navigation using the center of area of detected free space around the robot is described. The proposed method uses only part of detected free space in front of the robot to compute a partial center of area. It is then used to guide the robot in a path suitable for smooth and robust wandering in complex environments. A simple modification in the algorithm can make it useful for obstacle avoidance in reaching a stimulus goal. The proposed method is used in some examples of simulated experiments on map navigation and wandering and it is compared with standard wandering using Aria library from MobileRobots. Also some experiments in obstacle avoidance navigation to reach a stimulus goal are shown in different maps.

José Ramón Álvarez-Sánchez, Félix de la Paz Lépez, José Manuel Cuadra Troncoso, José Ignacio Rosado Sánchez

Mathematical Foundations of the Center of Area Method for Robot Navigation

The objective of this paper is to develop further the idea of using potential fields for robot navigation but changing to representation of free space instead of obstacles, because the task in robot navigation is to move in the free space not to identify the objects, and also by extending the methods to use directly virtual forces instead of the potentials as the base for robot movement, because not all driving forces will derive from a potential. After extending to a general virtual force we can select the simplest force to obtain a practical method to drive the robot by the center of area through safe places. Some particular cases of simple environments (straight wall, closed and open rooms, and corridor with a bend) are studied to analyze the properties of the proposed method, obtaining for them the resulting directions fields.

Félix de la Paz López, José Ramón Álvarez-Sánchez, José Ignacio Rosado Sánchez, José Manuel Cuadra Troncoso

Determining Sound Source Orientation from Source Directivity and Multi-microphone Recordings

This paper presents an analytic method for determining the orientation of a directional sound source in three-dimensional space using the source position, directivity and multi-microphone recordings. The acoustic signal emitted by the source is assumed to be broadband, such as a down-swept frequency modulated chirp of the kind many bats use while echolocating. The method has been tested in simulations on PC using the directivity of a piston transducer and the more complex and more realistic head-related transfer function of the

Phyllostomus discolor

bat. The ultimate purpose of the work is to determine the orientation and actual emitted call of a flying bat from a remote array recording.

Francesco Guarato, John C. T. Hallam

A Braitenberg Lizard: Continuous Phonotaxis with a Lizard Ear Model

The peripheral auditory system of a lizard is structured as a pressure difference receiver with strong broadband directional sensitivity. Previous work has demonstrated that this system can be implemented as a set of digital filters generated by considering the lumped-parameter model of the auditory system, and can be used successfully for step control steering of mobile robots. We extend the work to the continuous steering case, implementing the same model on a Braitenberg vehicle-like robot. The performance of the robot is evaluated in a phonotaxis task. The robot shows strong directional sensitivity and successful phonotaxis for a sound frequency range of 1400 Hz–1900 Hz. We conclude that the performance of the model in the continuous control task is comparable to that in the step control task.

Danish Shaikh, John Hallam, Jakob Christensen-Dalsgaard, Lei Zhang

A New Metric for Supervised dFasArt Based on Size-Dependent Scatter Matrices That Enhances Maneuver Prediction in Road Vehicles

In previous investigations, a supervised version of a dynamic FasArt method (SdFasArt) proved its capability to supply good results to the problem of maneuver prediction in road vehicles. The dynamic character of dFasArt minimized problems caused by noise in the sensors and provided stability on the predicted maneuvers. This paper presents a new SdFasArt architecture enhanced by the inclusion of size-dependent scatter matrices (SDSM) to compute the activation of the neurons. In this novel approach, the receptive fields of the neurons are capable to rotate and scale in order to better respond to data distributions with a preferred orientation in the input space, what leads to a more efficient classification. The results achieved by both methods in a series of experiments in real scenarios with a probe vehicle show that SDSM-SdFasArt supplies better results in terms of maneuver prediction and number of nodes.

Ana Toledo, Rafael Toledo-Moreo, José Manuel Cano-Izquierdo

A Strategy for Evolutionary Spanning Tree Construction within Constrained Graphs with Application to Electrical Networks

In this work we present a particular encoding and fitness evaluation strategy for a genetic approach in the context of searching in graphs. In particular, we search for a spanning tree in the universe of directed graphs under certain constraints related to the topology of the graphs considered. The algorithm was also implemented and tested as a new topological approach to electrical power network observability analysis and was revealed as a valid technique to manage observability analysis when the system is unobservable. The algorithm was tested on benchmark systems as well as on networks of realistic dimensions.

Santiago Vazquez-Rodriguez, Richard J. Duro

An Evolutionary Approach for Correcting Random Amplified Polymorphism DNA Images

Random amplified polymorphism DNA (RAPD) analysis is a widely used technique in studying genetic relationships between individuals, in which processing the underlying images is a quite difficult problem, affected by various factors. Among these factors, noise and distortion affect the quality of images, and subsequently, accuracy in interpreting the data. We propose a method for processing RAPD images that allows to improve their quality and thereof, augmenting biological conclusions. This work presents a twofold objective that attacks the problem by considering two noise sources: band distortion and lane misalignment in the images. Genetic algorithms have shown good results in treating difficult problems, and the results obtained by using them in this particular problem support these directions for future work.

M. Angélica Pinninghoff J., Ricardo Contreras A., Luis Rueda

A Method to Minimize Distributed PSO Algorithm Execution Time in Grid Computer Environment

This paper introduces a method to minimize distributed PSO algorithm execution time in a grid computer environment, based on a reduction in the information interchanged among the demes involved in the process of finding the best global fitness solution. Demes usually interchange the best global fitness solution they found at each iteration. Instead of this, we propose to interchange information only after an specified number of iterations are concluded. By applying this technique, it is possible to get a very significant execution time decrease without any loss of solution quality.

F. Parra, S. Garcia Galan, A. J. Yuste, R. P. Prado, J. E. Muñoz

Assessment of a Speaker Recognition System Based on an Auditory Model and Neural Nets

This paper deals with a new speaker recognition system based on a model of the human auditory system. Our model is based on a human nonlinear cochlear filter-bank and Neural Nets.

The efficiency of this system has been tested using a number of Spanish words from the ‘Ahumada’ database as uttered by a native male speaker. These words were fed into the cochlea model and their corresponding outputs were processed with an envelope component extractor, yielding five parameters that convey different auditory sensations (loudness, roughness and virtual tones).

Because this process generates large data sets, the use of multivariate statistical methods and Neural Nets was appropriate. A variety of normalization techniques and classifying methods were tested on this biologically motivated feature set.

Ernesto A. Martínez–Rams, Vicente Garcerán–Hernández

CIE-9-MC Code Classification with knn and SVM

This paper is concerned with automatic classification of texts in a medical domain. The process consists in classifying reports of medical discharges into classes defined by the

CIE-9-MC

codes. We will assign

CIE-9-MC

codes to reports using either a

knn

model or support vector machines. One of the added values of this work is the construction of the collection using the discharge reports of a medical service. This is a difficult collection because of the high number of classes and the uneven balance between classes. In this work we study different representations of the collection, different classication models, and different weighting schemes to assign

CIE-9-MC

codes. Our use of document expansion is particularly novel: the training documents are expanded with the descriptions of the assigned codes taken from

CIE-9-MC

. We also apply SVMs to produce a ranking of classes for each test document. This innovative use of SVM offers good results in such a complicated domain.

David Lojo, David E. Losada, Álvaro Barreiro

Time Estimation in Injection Molding Production for Automotive Industry Based on SVR and RBF

Resource planning in automotive industry is a very complex process which involves the management of material and human needs and supplies. This paper deals with the production of plastic injection moulds used to make car components in the automotive industry. An efficient planning requires, among other, an accurate estimation of the task execution times in the mould production process. If the relation between task times and mould parts geometry is known, the moulds can be designed with a geometry that allows the shortest production time. We applied two popular regression approaches, Support Vector Regression and Radial Basis Function, to this problem, achieving accurate results which make feasible an automatic estimation of the task execution time.

M. Reboreda, M. Fernández-Delgado, S. Barro

Performance of High School Students in Learning Math: A Neural Network Approach

This paper depicts a research work that uses neural networks to predict academic performance in mathematics, focusing on students enrolled in a public school in Chile. This proposal identifies social, knowledge and psychological issues that impact upon successful learning in a meaningful way. The experience considers different instruments used to gather the necessary information for training the neural network. This information includes the level of knowledge, the logical-mathematical intelligence, the students’ self-esteem and about 80 factors considered as relevant in an international project known as PISA. The most adequate network configuration can be found with different experiments. Results show a good predictive level and point out the importance of using local data for fine tuning.

Ricardo Contreras A., Pedro Salcedo L., M. Angélica Pinninghoff J.

Backmatter

Weitere Informationen