• Introduction of CausalMorph , a three-stage algorithm that projects observational data toward the regime assumed by Linear Non-Gaussian Acyclic Models (LiNGAM). • Application of CausalMorph to 17,280 unique data configurations results in a significant 37.7% relative reduction in Structural Hamming Distance (SHD) for downstream DirectLiNGAM ( p < .001). • Evidience of a regularization effect is found, with improved causal discovery accuracy even under ideal LiNGAM conditions, indicating mitigation of finite-sample artifacts. • Provides evidence for data projection as a practical strategy for extending the applicability of the LiNGAM framework. Moving from associative learning to inferring cause–effect relationships remains a central challenge for intelligent systems. The Linear Non-Gaussian Acyclic Model (LiNGAM) family identifies a single, fully directed causal graph from observational data rather than an equivalence class. However, deviations from its assumptions of linearity and non-Gaussian noise limit its applicability. To address this, this paper introduces CausalMorph, a data preconditioning algorithm that projects observational data toward a regime compatible with LiNGAM. The projection employs a three-stage sequence: local linearization of causal mechanisms, synthesis of non-Gaussian residuals, and orthogonalization of parent-residual dependencies. Across an evaluation of 34,560 synthetic paired experiments, CausalMorph yielded significant reductions in Structural Hamming Distance (SHD) of 37.7% ± 10.8% for DirectLiNGAM and 16.4% ± 13.8% for ICA-LiNGAM ( p < 0.001). Additionally, the CausalMorph + DirectLiNGAM pipeline achieved a lower mean SHD than the differentiable non-linear baseline algorithm in both linear and non-linear regimes. By operating as a non-iterative, single-pass projection, the method avoids the k iter optimization loops required by continuous frameworks, offering a highly efficient path to structural recovery. The algorithm also systematically rescues baseline solvers from catastrophic large-sample traps under fully Gaussian noise, and maintains an 85.8% win rate over the baseline when utilizing an autonomous data-driven initialization for the prior causal order. These findings suggest statistical projection as a viable and structurally conservative strategy for applying LiNGAM-based causal discovery to data environments that violate its base assumptions.

This research presents an affective embodied agent to generate empathic interactions in an upper limb virtual rehabilitation platform, enhancing the platform's usability and encouraging its use at home without close supervision by physiotherapists. The agent interacts using natural language (Spanish) and simulating affective states through facial expressions and voice intonation. An interaction component set was defined by considering the opinions of several physiotherapists, including an affective state set (neutrality, enthusiasm, happiness, surprise) and a set of phrases to be expressed by the agent. A 3D model was designed and since affective state perception is subjective, its facial expressions were chosen through a consensus-based study. The agent's affective behavior was defined by a control model that receives the patient's affective state and performance as inputs. It is a rule-based system with averaging and synchronization techniques and a selection method based on probability distributions to provide more natural interactions. A usability evaluation was performed on both versions of the platform, with and without the agent, on a group of 12 non-disabled participants and another of 12 disabled. The results suggest that the agent provides empathic and motivating user experiences, enhances usability, and encourages the use of the platform at home without close expert supervision

As a safety-critical application, automated driving aims to provide autonomy and safe navigation. Although recent advances in perception algorithms based on deep learning have boosted the progress in such field, state of the art depends on availability of large datasets. Thus, most visual analysis in this context is related to the detection and classification of previously known classes. However, dynamic environments with complex interactions among traffic elements can lead to unpredictable anomalous situations that are not registered previously. On the other hand, most of prior work on visual novelty detection points out novelty instances but does not provide additional useful information that can be used in the next levels for decision-making. In this paper we propose an approach to detect global visual novelties and their corresponding type in real driving scenarios. Based on pixel-wise and perceptual information, we use a generative adversarial network to detect novelties and a support vector machine to identify their categories. Its performance is experimentally evaluated using the Ford self-driving dataset. Experimental results show that the average area under the curve (AUC) surpasses 0.97 for novelty detection, and novelty type identification reaches an average of 92.7% of accuracy requiring only a small number of samples for training.

Computational systems that process multiple affective states may benefit from explicitly considering the interaction between the states to enhance their recognition performance. This work proposes the combination of a multi-label classifier, Circular Classifier Chain (CCC), with a multimodal classifier, Fusion using a Semi-Naive Bayesian classifier (FSNBC), to include explicitly the dependencies between multiple affective states during the automatic recognition process. This combination of classifiers is applied to a virtual rehabilitation context of post-stroke patients. We collected data from post-stroke patients, which include finger pressure, hand movements, and facial expressions during ten longitudinal sessions. Videos of the sessions were labelled by clinicians to recognize four states: tiredness, anxiety, pain, and engagement. Each state was modelled by the FSNBC receiving the information of finger pressure, hand movements, and facial expressions. The four FSNBCs were linked in the CCC to exploit the dependency relationships between the states. The convergence of CCC was reached by 5 iterations at most for all the patients. Results (ROC AUC) of CCC with the FSNBC are over$0.940 \pm 0.045$($mean \pm std.\;deviation$) for the four states. Relationships of mutual exclusion between engagement and all the other states and co-occurrences between pain and anxiety were detected and discussed.

Data from multiple sensors can boost the automatic recognition of multiple affective states in a multilabel and multimodal recognition system. At any time, the streaming from any of the contributing sensors can be missing. This work proposes a method for dealing with a missing sensor in a multilabel and multimodal automatic affective states recognition system. The proposed method, called Hot Deck using Conditional Probability Tables (HD-CPT), is incorporated into a multimodal affective state recognition system for compensating the loss of a sensor using the recorded historical information of the sensor and its interaction with the other available sensors. In this work, we consider a multilabel classifier, named Circular Classifier Chain, for the automatic recognition of four states: tiredness, anxiety, pain, and engagement; combined with a multimodal classifier based on three sensors: fingers pressure, hand movements, and facial expressions; which was adapted for coping with the problem of a missing sensor in a virtual rehabilitation platform for post-stroke patients. A dataset of five post-stroke patients who attended ten longitudinal rehabilitation sessions was used for the evaluation. The inclusion of HD-CPT compensated for the loss of one sensor with results above those obtained with only the remaining sensors available. HD-CPT prevents the system from collapsing when a sensor fails, providing continuity of operation with results that attenuate the loss of the sensor. The proposed method HD-CPT can provide robustness for the naturalistic everyday use of an affective states recognition system.

Stress assessment is a complex issue and numerous studies have examined factors that influence stress in working environments. Research studies have shown that monitoring individuals' behaviour parameters during daily life can also help assess stress levels. In this study, we examine assessment of work-related stress using features derived from sensors in smartphones. In particular, we use information from physical activity levels, location, social-interactions, social-activity, and application usage during working days. Our study included 30 employees chosen from two different private companies, monitored over a period of 8 weeks in real work environments. The findings suggest that information from phone sensors shows important correlation with employees perceived stress level. Second, we used machine learning methods to classify perceived stress levels based on the analysis of information provided by smartphones. We used decision trees obtaining 67.57 percent accuracy and 71.73 percent after applying a semi-supervised method. Our results show that stress levels can be monitored in unobtrusive manner, through analysis of smartphone data.

Virtual rehabilitation environments may afford greater patient personalization if they could harness the patient's affective state. Four states: anxiety, pain, engagement and tiredness (either physical or psychological), were hypothesized to be inferable from observable metrics of hand location and gripping strength-relevant for rehabilitation. Contributions are; (a) multiresolution classifier built from Semi-Naïve Bayesian classifiers, and (b) establishing predictive relations for the considered states from the motor proxies capitalizing on the proposed classifier with recognition levels sufficient for exploitation. 3D hand locations and gripping strength streams were recorded from 5 post-stroke patients whilst undergoing motor rehabilitation therapy administered through virtual rehabilitation along 10 sessions over 4 weeks. Features from the streams characterized the motor dynamics, while spontaneous manifestations of the states were labelled from concomitant videos by experts for supervised classification. The new classifier was compared against baseline support vector machine (SVM) and random forest (RF) with all three exhibiting comparable performances. Inference of the aforementioned states departing from chosen motor surrogates appears feasible, expediting increased personalization of virtual motor neurorehabilitation therapies.

The automatic recognition of multiple affective states can be enhanced if the underpinning computational models explicitly consider the interactions between the states. This work proposes a computational model that incorporates the dependencies between four states (tiredness, anxiety, pain, and engagement)known to appear in virtual rehabilitation sessions of post-stroke patients, to improve the automatic recognition of the patients' states. A dataset of five stroke patients which includes their fingers' pressure (PRE), hand movements (MOV)and facial expressions (FAE)during ten sessions of virtual rehabilitation was used. Our computational proposal uses the Semi-Naive Bayesian classifier (SNBC)as base classifier in a multiresolution approach to create a multimodal model with the three sensors (PRE, MOV, and FAE)with late fusion using SNBC (FSNB classifier). There is a FSNB classifier for each state, and they are linked in a circular classifier chain (CCC)to exploit the dependency relationships between the states. Results of CCC are over 90% of ROC AUC for the four states. Relationships of mutual exclusion between engagement and all the other states and some co-occurrences between pain and anxiety for the five patients were detected. Virtual rehabilitation platforms that incorporate the automatic recognition of multiple patient's states could leverage intelligent and empathic interactions to promote adherence to rehabilitation exercises.

Automatic image annotation and image segmentation are two prominent research fields of Computer Vision, that are getting higher attention these days to accomplish image analysis and scene understanding. In this work, we present an annotation algorithm based on a hierarchical image partition, that makes use of Markov Random Fields (MRFs) to model spatial and hierarchical relations among regions in the image. In this way, we can capture local, global and contextual information. Also, we combine the processes of annotation and segmentation in an iterative way so that each process benefits from the other. Furthermore, we investigate the selection of the starting segmentation level for the hierarchical annotation process, to show its relevance for the final results. We experimentally validate our approach in three well-known datasets: CorelA, Stanford Background and MSRC-21 datasets. In these datasets, we achieved better or comparable results to other state-of-the-art algorithms, improving our base classifier in all cases.

This paper presents a new foveation-based method in the discrete cosine transform (DCT) domain to preserve the privacy rights of subjects through face de-identification while preserving awareness for gender and facial expression classification tasks. A comparative analysis between the commonly used ad-hoc methods for image obfuscation and the proposed method is performed. The awareness-privacy tradeoff at different obfuscation levels is quantified by using a support vector machine (SVM) classifier and a two-directional two-dimensional Principal Component Analysis method. Experimental results using a subset of the FERET database show that the new technique accomplishes higher gender and facial expression classification rates at lower face recognition rates.

One of the main problems of recognizing faces in videos is to achieve accurate algorithms which can be used in real-time applications. Recently, Fisher Vector representation of local descriptors (e.g., SIFT) has gained widespread popularity, achieving good recognition rates. In this work, we propose to use Fisher Vector encoding of binary features for video face recognition, in order to speed up the computation time of the representation. The experimental evaluation was conducted on the challenging YouTube Faces database, showing that the proposed method is very efficient, and has an accuracy comparable with state-of-the-art methods.

In this work an invariant shape features extraction, description and matching method (LISF) is proposed. In order to balance the discriminative power and the robustness to noise and occlusion in the contour, local features are extracted from contour to describe shape, which are later matched globally. The proposed extraction, description and matching methods are invariant to rotation, translation, and scale and present certain robustness to partial occlusion. Its invariability and robustness are validated by the performed experiments in shape retrieval and classification tasks. Experiments were carried out in the Shape99, Shape216, and MPEG-7 datasets, where different artifacts were artificially added to obtain partial occlusion as high as 60%. For the highest occlusion levels the proposed method outperformed other popular shape description methods, with about 20% higher bull's eye score and 25% higher accuracy in classification.

The task addressed in this paper is to plan iteratively a set views in order to reconstruct an object using a mobile manipulator robot with an “eye-in-hand” sensor. The proposed method plans views directly in the configuration space avoiding the need of inverse kinematics. It is based on a fast evaluation and rejection of a set of candidate configurations. The main contributions are: a utility function to rank the views and an evaluation strategy implemented as a series of filters. Given that the candidate views are configurations, motion planning is solved using a rapidly-exploring random tree. The system is experimentally evaluated in simulation, contrasting it with previous work. We also present experiments with a real mobile manipulator robot, demonstrating the effectiveness of our method.

Intelligent agents, human or artificial, often change their behaviour as they interact with other agents. For an agent to optimise its performance when interacting with such agents, it must be capable of detecting and adapting according to such changes. This work presents an approach on how to effectively deal with non-stationary switching opponents in a repeated game context. Our main contribution is a framework for online learning and planning against opponents that switch strategies. We present how two opponent modelling techniques work within the framework and prove the usefulness of the approach experimentally in the iterated prisoner's dilemma, when the opponent is modelled as an agent that switches between different strategies (e.g. TFT, Pavlov and Bully). The results of both models were compared against each other and against a state-of-the-art non-stationary reinforcement learning technique. Results reflect that our approach obtains competitive results without needing an offline training phase, as opposed to the state-of-the-art techniques.

Following the increase in the different areas of knowledge of multimedia databases, image retrieval by content (CBIR) has captured attention in the last decade. In this paper we present an image retrieval system of art, developed to query and exploration of a collection of art images by QBE, considering similarity criteria: the color palette, the use of light and the technique used. These similarity criteria are analyzed by extracting low-level features such as color, texture and spatial relationships.

Intelligent environments need to acquire, combine and interpret the user's requests and take the best decisions according to the user needs. Thus, they require intelligent agents that reason under uncertainty to achieve the system goals. Probabilistic graphical models (PGMs) allow intelligent agents to reason and take optimal decisions under uncertainty, in an effective and efficient way. We present an overview of PGMs and describe two applications for intelligent environments: (i) information validation, (ii) adaptation to the user.

Abstract This paper introduces a novel ranking refinement approach based on relevance feedback for the task of document retrieval. We focus on the problem of ranking refinement since recent evaluation results from Information Retrieval (IR) systems indicate that current methods are effective retrieving most of the relevant documents for different sets of queries, but they have severe difficulties to generate a pertinent ranking of them. Motivated by these results, we propose a novel method to re-rank the list of documents returned by an IR system. The proposed method is based on a Markov Random Field (MRF) model that classifies the retrieved documents as relevant or irrelevant. The proposed MRF combines: (i) information provided by the base IR system, (ii) similarities among documents in the retrieved list, and (iii) relevance feedback information. Thus, the problem of ranking refinement is reduced to that of minimising an energy function that represents a trade-off between document relevance and inter-document similarity. Experiments were conducted using resources from four different tasks of the Cross Language Evaluation Forum (CLEF) forum as well as from one task of the Text Retrieval Conference (TREC) forum. The obtained results show the feasibility of the method for re-ranking documents in IR and also depict an improvement in mean average precision compared to a state of the art retrieval machine.

In multidimensional classification the goal is to assign an instance to a set of different classes. This task is normally addressed either by defining a compound class variable with all the possible combinations of classes (label power-set methods, LPMs) or by building independent classifiers for each class (binary-relevance methods, BRMs). However, LPMs do not scale well and BRMs ignore the dependency relations between classes. We introduce a method for chaining binary Bayesian classifiers that combines the strengths of classifier chains and Bayesian networks for multidimensional classification. The method consists of two phases. In the first phase, a Bayesian network (BN)that represents the dependency relations between the class variables is learned from data. In the second phase, several chain classifiers are built, such that the order of the class variables in the chain is consistent with the class BN. At the end we combine the results of the different generated orders. Our method considers the dependencies between class variables and takes advantage of the conditional independence relations to build simplified models. We perform experiments with a chain of naïve Bayes classifiers on different benchmark multidimensional datasets and show that our approach out performs other state-of-the-art methods.

This paper elaborates on the benefits of using particle swarm model selection (PSMS) for building effective ensemble classification models. PSMS searches in a toolbox for the best combination of methods for preprocessing, feature selection and classification for generic binary classification tasks. Throughout the search process PSMS evaluates a wide variety of models, from which a single solution (i.e. the best classification model) is selected. Satisfactory results have been reported with the latter formulation in several domains. However, many models that are potentially useful for classification are disregarded for the final model. In this paper we propose to re-use such candidate models for building effective ensemble classifiers. We explore three simple formulations for building ensembles from intermediate PSMS solutions that do not require of further computation than that of the traditional PSMS implementation. We report experimental results on benchmark data as well as on a data set from object recognition. Our results show that better models can be obtained with the ensemble version of PSMS, motivating further research on the combination of candidate PSMS models. Additionally, we analyze the diversity of the classification models, which is known to be an important factor for the construction of ensembles.

Markov decision processes (MDPs) have become a standard method for planning under uncertainty, however they usually assume a sequential process, so a single action is executed at each time step. In some applications, as in robotics, it is required to execute several actions concurrently. For this we propose a framework based on a functional decomposition of the problem into several sub-problems, each represented as a subMDP. Each subMDP is solved independently and their policies are combined to obtain a global solution, such that the actions of each subMDP can be executed concurrently. As we combine the local policies, conflicts between them can arise. We define two kinds of conflicts, resource and behavior conflicts, and propose solutions for both. Resource conflicts are solved off-line via a two-phase process which guarantees a near-optimal global policy. Behavior conflicts are solved on-line based on a set of restrictions specified by the user. If there are no restrictions, all the actions are executed concurrently; otherwise, an arbiter selects the action(s) with higher expected utility. We present experimental results in two cases: (i) a simulated robot navigation problem, with resource conflicts, and (ii) a simulated robot in a message delivery task, with behavior conflicts.

These days researchers are looking for design and development of autonomous robots. It is desirable that robots be capable to acquire the information they need to perform actions and make decisions. Any mobile or humanoid robot will need to construct a spatial representation or model of the surrounding environment that allows it to move and execute tasks with success. The reconstruction of an environment is an important and useful capability for these kind of robots. In order to construct a model, the robot needs to obtain information through a series of acquisitions from its sensors by solving occlusions. Therefore, an important issue is how to plan these robot placements (views) optimally, according to certain criteria for the purpose of reconstructing a complete model automatically. In this work we present a view planning algorithm to solve the problem of 3D modelling for indoor environments; the algorithm uses a volumetric representation as a reasoning domain. In this paper we propose the use of probability distribution functions as a model for the desirable behavior of the system, considering perception range data. The method uses a maximum a posteriori estimator to find the perception system parameters that defines the next best view position. We present results in simulation for a five degrees of freedom robot with a 3D range camera mounted on it to validate our approach.

For manipulating an unknown object a robot needs a 3D model of it. Given the limited field of view of a camera and self occlusions, a set of views is required to build a complete 3D model, so an important problem is how to select these views optimally according to certain criteria. We propose a novel algorithm to select the next-best-view (NBV) for a range camera to model 3D arbitrary objects. We use a volumetric representation and voxel labeling. We propose a new utility function based on factors considering voxel, quality and navigation information. We also propose two novel strategies to make faster the search of the NBV, one based on a hierarchical decomposition of the search space and other based on a multi-resolution of ray tracing. We have tested our planner in simulation with 7 different 3D objects, showing good results in terms of quality of the models and computation time required, and at the same time reducing the distance that the sensor has to travel to obtain the set of views.

The ability to recognize people is a key element for improving human-robot interaction in service robots. There are many approaches for face recognition; however, these assume unrealistic conditions for a service robot, like having an image with a centered face under controlled illumination. We have developed a novel face recognition system so that a mobile robot can learn new faces and recognize them in real-time in realistic indoor environments. It is able to learn online a new face based on a single frame, which is later used to recognize the person even under different environmental conditions. We employ a preprocessing step to reduce the effect of different illumination conditions, and then identify 3 regions in the face: left eye, right eye and nose-mouth. SIFT features are extracted from each region and stored in a feature vector, which is used for recognition. The matching strategy is able to discard unknown faces and the recognition process uses a Bayesian approach over several frames to improve accuracy. Experimental results in natural environment and with Yale's face database show that that our approach is able to learn different faces from a single image and recognize them on average in three seconds, with very competitive results.

We propose the use of hidden Markov models (HMMs) to qualify arm gestures. A HMM is trained based on the reference or correct gesture. Then, samples of the gesture that we want to score are used to train a second HMM. Both HMMs are compared, and a measure of their similarity is used to qualify the gesture. We used 3 different metrics to compare HMMs: Levinson, Kullback-Leibler and Porikli. For this, a visual system was developed to track a person's arm, which serves as input to the models that qualify the gestures. We applied this method to qualify the arm movements of stroke patients under rehabilitation. We analyzed three therapeutic gestures: flexion, circular and abduction. A HMM is trained to represent the movement of a healthy person for each gesture, which is compared with the HMMs obtained for each patient. The results are compared with the scales that are used in therapy. From the analysis of several experiments, the Porikli metric was the best to qualify the three gestures, in terms of the motricity index.

We have developed an affective behavior model (ABM) for intelligent tutoring systems, so the tutor considers the affective state as well as the pedagogical state of the student. The ABM first establishes the affective state based on the OCC cognitive model of emotions, represented as a dynamic Bayesian network. Then, the tutor selects the best tutorial action according to the pedagogical and affective state of the student. For this it uses a dynamic decision network that was designed based on the opinions of a group of experienced teachers. The ABM has been integrated to an intelligent tutor for mobile robotics. A pilot study with a group of 20 students shows that the affective and tutorial actions selected by the tutor are in general in agreement with the students’ expectations.

In this paper a novel approach for improving automatic image annotation methods is proposed. The approach is based on the fact that accuracy of current image annotation methods is low if we look at the most confident label only. Instead, accuracy is improved if we look for the correct label within the set of the top−k candidate labels. We take advantage of this fact and propose a Markov random field (MRF) based on word co-occurrence information for the improvement of annotation systems. Through the MRF structure we take into account spatial dependencies between connected regions. As a result, we are considering semantic relationships between labels. We performed experiments with iterated conditional modes and simulated annealing as optimization strategies in a subset of the Corel benchmark collection. Experimental results of the proposed method together with a k−nearest neighbors classifier as our annotation method show important error reductions. 1

Artificial neural networks have been proven to be effective learning algorithms since their introduction. These methods have been widely used in many domains, including scientific, medical, and commercial applications with great success. However, selecting the optimal combination of preprocessing methods and hyperparameters for a given data set is still a challenge. Recently a method for supervised learning model selection has been proposed: Particle Swarm Model Selection (PSMS). PSMS is a reliable method for the selection of optimal learning algorithms together with preprocessing methods, as well as for hyperparameter optimization. In this paper we applied PSMS for the selection of the (pseudo) optimal combination of preprocessing methods and hyperparameters for a fixed neural network on benchmark data sets from a challenging competition: the (IJCNN 2007) agnostic vs prior knowledge challenge. A forum for the evaluation of methods for model selection and data representation discovery. In this paper we further show that the use of PSMS is useful for model selection when we have no knowledge about the domain we are dealing with. With PSMS we obtained competitive models that are ranked high in the official results of the challenge.

During the operation of complex dynamic systems, the operator is exposed to unusual and abnormal situations, and has to be able to discriminate erroneous inputs, and promptly identify the source of the problem in order to define the corrective actions to be taken. In this work, the development and evaluation of an Intelligent Assistant for Operator’s Training (IAOT) is presented. It uses, as input, a recommended plan generated by a decision-theoretic planner, which establishes the sequence of actions that will allow a steam generator to reach its optimum operation. The system monitors the operator's actions and detects his/her discrepancies, compared to those recommended by the planner. On the basis of the above, follow-up advice is given on-line, and an evaluation of performance is given later. Results obtained from an experimental evaluation of IAOT show that the usage of the intelligent assistant can be helpful to increase operator performance.

The successful performance of machining centers involves selection, control and monitoring of a large number of parameters. Cutting tool condition is one of the most important variables due to the strong influence on dimensional accuracy and surface finish of the product. However, the cutting tool condition is difficult to measure online. A new proposal for online monitoring of the cutting tool condition based on Hidden Markov Models is presented. The generated vibration signals between the cutting tool and the workpiece are used for classification of the tool condition in different states (new, half-new, half-worn, and worn cutting tool). Feature vectors were obtained from these vibration signals by applying a triangular filter bank and computing the Mel Frequency Cepstrum Coefficients. The proposal includes a training step using the Baum-Welch algorithm and a diagnosis step exploiting the Viterbi algorithm. A Monte Carlo simulation validates a successful performance using experimental data in a face milling process. A comparison with classical approaches such as Artificial Neural Network is discussed.

This paper develops a new theory and model for information and sensor validation. The model represents relationships between variables using Bayesian networks and utilizes probabilistic propagation to estimate the expected values of variables. If the estimated value of a variable differs from the actual value, an apparent fault is detected. The fault is only apparent since it may be that the estimated value is itself based on faulty data. The theory extends our understanding of when it is possible to isolate real faults from potential faults and supports the development of an algorithm that is capable of isolating real faults without deferring the problem to the use of expert provided domain-specific rules. To enable practical adoption for real-time processes, an any time version of the algorithm is developed, that, unlike most other algorithms, is capable of returning improving assessments of the validity of the sensors as it accumulates more evidence with time. The developed model is tested by applying it to the validation of temperature sensors during the start-up phase of a gas turbine when conditions are not stable; a problem that is known to be challenging. The paper concludes with a discussion of the practical applicability and scalability of the model.

We propose a visual clue for navigation, in which the local 3D information or (p, q) is summarised in a single structure, called the pq-histogram. The pq-histogram is obtained by discretising in logarithmic scale the local depth (p, q). Since this histogram gives a global view of the scene, it could be used as a fast clue for navigation. An arrangement of four pq-histograms, one for each image quadrant, can advise many orientations. The potential application of the pq-histogram is illustrated in two domains. Firstly, semi-automatic navigation of an endoscope inside the human colon. Secondly, mobile robot navigating in certain environments, such as corridors and mines. In both cases, the method was tested with real images with encouraging results.

A method for exploring and navigating autonomously in indoor environments is described. This method merges a local strategy, similar to wall following to keep the robot close to obstacles, within a global search frame, based on a dynamic programming algorithm. We introduce the concept of travel space as a way to map costs to grid cells based on distances to obstacles. This hybrid approach takes advantages of local strategies that consider perceptual limitations of sensors without losing the completeness of a global search. This exploration and navigation method is tested using a simulated and a real mobile robot with promising results.

We present a probabilistic approach for local scale selection. The proposed method computes a probability measure on scale space, which is based on Bayesian estimation theory, and it leads to an efficient computational implementation. At each scale we associate a decomposition likelihood, one for the smoothed image and other for the residual. The scale selection method, based on the minimal description length principle, maximizes the likelihood of the observed image given the local scale, and at the same time, minimizes the residual. Initial experiments show that our approach can be successfully applied to edge detection, and also to adaptive Gaussian filtering and texture segmentation.

A novel approach for human silhouette recognition is presented. The method is based on Fourier descriptors. We made an analysis of which and how many descriptors are enough to have a general human silhouette representation, and concluded that a reduced number of components, low and high frequency, is sufficient for representing a human silhouette and for its recognition in different poses. Based on this study, we developed a system that uses 40 normalized descriptors and a nearest centroid classified for human silhouette recognition. The method was tested with real images of humans and other objects with similar contours, achieving a 97% correct recognition.

A method for object recognition that is invariant under translation, rotation and scaling is addressed. The first step of the method (pre-processing) takes into account the invariant properties of the normalized moment of inertia and a novel coding that extracts topological object characteristics. The second step (recognition) is achieved by using a holographic nearest-neighbor (HHN) algorithm, in which vectors obtained in the pre-processing step are used as inputs to it. The algorithm is tested in character recognition, using the 26 upper-case letters of the alphabet. Only four different orientations and one size (for each letter) were used for training. Recognition was tested with 17 different sizes and 14 rotations. The results are encouraging, since we achieved 98% correct recognition. Tolerance to boundary deformations and random noise was tested. Results for character recognition in "real" images of car plates are presented as well.