Reconhecimento de língua gestual

Trabalho Final de Mestrado para obtenção do grau de Mestre em Engenharia Informática e Computadores

Pré-processamento e Tag automático de imagens em ambiente móvel e web, aplicado a um sistema de informação geográfica

No panorama socioeconómico atual, a contenção de despesas e o corte no financiamento de serviços secundários consumidores de recursos conduzem à reformulação de processos e métodos das instituições públicas, que procuram manter a qualidade de vida dos seus cidadãos através de programas que se mostrem mais eficientes e económicos. O crescimento sustentado das tecnologias móveis, em conjunção com o aparecimento de novos paradigmas de interação pessoa-máquina com recurso a sensores e sistemas conscientes do contexto, criaram oportunidades de negócio na área do desenvolvimento de aplicações com vertente cívica para indivíduos e empresas, sensibilizando-os para a disponibilização de serviços orientados ao cidadão. Estas oportunidades de negócio incitaram a equipa do projeto a desenvolver uma plataforma de notificação de problemas urbanos baseada no seu sistema de informação geográfico para entidades municipais. O objetivo principal desta investigação foca a idealização, conceção e implementação de uma solução completa de notificação de problemas urbanos de caráter não urgente, distinta da concorrência pela facilidade com que os cidadãos são capazes de reportar situações que condicionam o seu dia-a-dia. Para alcançar esta distinção da restante oferta, foram realizados diversos estudos para determinar características inovadoras a implementar, assim como todas as funcionalidades base expectáveis neste tipo de sistemas. Esses estudos determinaram a implementação de técnicas de demarcação manual das zonas problemáticas e reconhecimento automático do tipo de problema reportado nas imagens, ambas desenvolvidas no âmbito deste projeto. Para a correta implementação dos módulos de demarcação e reconhecimento de imagem, foram feitos levantamentos do estado da arte destas áreas, fundamentando a escolha de métodos e tecnologias a integrar no projeto. Neste contexto, serão apresentadas em detalhe as várias fases que constituíram o processo de desenvolvimento da plataforma, desde a fase de estudo e comparação de ferramentas, metodologias, e técnicas para cada um dos conceitos abordados, passando pela proposta de um modelo de resolução, até à descrição pormenorizada dos algoritmos implementados. Por último, é realizada uma avaliação de desempenho ao par algoritmo/classificador desenvolvido, através da definição de métricas que estimam o sucesso ou insucesso do classificador de objetos. A avaliação é feita com base num conjunto de imagens de teste, recolhidas manualmente em plataformas públicas de notificação de problemas, confrontando os resultados obtidos pelo algoritmo com os resultados esperados.

A Hybrid Architecture for Recognising Speech Signals in Malayalam

Speech is the primary, most prominent and convenient means of communication in audible language. Through speech, people can express their thoughts, feelings or perceptions by the articulation of words. Human speech is a complex signal which is non stationary in nature. It consists of immensely rich information about the words spoken, accent, attitude of the speaker, expression, intention, sex, emotion as well as style. The main objective of Automatic Speech Recognition (ASR) is to identify whatever people speak by means of computer algorithms. This enables people to communicate with a computer in a natural spoken language. Automatic recognition of speech by machines has been one of the most exciting, significant and challenging areas of research in the field of signal processing over the past five to six decades. Despite the developments and intensive research done in this area, the performance of ASR is still lower than that of speech recognition by humans and is yet to achieve a completely reliable performance level. The main objective of this thesis is to develop an efficient speech recognition system for recognising speaker independent isolated words in Malayalam.

Development of thermo-analytical prediction and classification models for food in thermal devices using a multi sensor system and artificial neural networks

Die thermische Verarbeitung von Lebensmitteln beeinflusst deren Qualität und ernährungsphysiologischen Eigenschaften. Im Haushalt ist die Überwachung der Temperatur innerhalb des Lebensmittels sehr schwierig. Zudem ist das Wissen über optimale Temperatur- und Zeitparameter für die verschiedenen Speisen oft unzureichend. Die optimale Steuerung der thermischen Zubereitung ist maßgeblich abhängig von der Art des Lebensmittels und der äußeren und inneren Temperatureinwirkung während des Garvorgangs. Das Ziel der Arbeiten war die Entwicklung eines automatischen Backofens, der in der Lage ist, die Art des Lebensmittels zu erkennen und die Temperatur im Inneren des Lebensmittels während des Backens zu errechnen. Die für die Temperaturberechnung benötigten Daten wurden mit mehreren Sensoren erfasst. Hierzu kam ein Infrarotthermometer, ein Infrarotabstandssensor, eine Kamera, ein Temperatursensor und ein Lambdasonde innerhalb des Ofens zum Einsatz. Ferner wurden eine Wägezelle, ein Strom- sowie Spannungs-Sensor und ein Temperatursensor außerhalb des Ofens genutzt. Die während der Aufheizphase aufgenommen Datensätze ermöglichten das Training mehrerer künstlicher neuronaler Netze, die die verschiedenen Lebensmittel in die entsprechenden Kategorien einordnen konnten, um so das optimale Backprogram auszuwählen. Zur Abschätzung der thermische Diffusivität der Nahrung, die von der Zusammensetzung (Kohlenhydrate, Fett, Protein, Wasser) abhängt, wurden mehrere künstliche neuronale Netze trainiert. Mit Ausnahme des Fettanteils der Lebensmittel konnten alle Komponenten durch verschiedene KNNs mit einem Maximum von 8 versteckten Neuronen ausreichend genau abgeschätzt werden um auf deren Grundlage die Temperatur im inneren des Lebensmittels zu berechnen. Die durchgeführte Arbeit zeigt, dass mit Hilfe verschiedenster Sensoren zur direkten beziehungsweise indirekten Messung der äußeren Eigenschaften der Lebensmittel sowie KNNs für die Kategorisierung und Abschätzung der Lebensmittelzusammensetzung die automatische Erkennung und Berechnung der inneren Temperatur von verschiedensten Lebensmitteln möglich ist.

Algorithmic statistical analysis of electrophysiological data for the investigation of structure-activity relationship in single neurons

We are developing computational tools supporting the detailed analysis of the dependence of neural electrophysiological response on dendritic morphology. We approach this problem by combining simulations of faithful models of neurons (experimental real life morphological data with known models of channel kinetics) with algorithmic extraction of morphological and physiological parameters and statistical analysis. In this paper, we present the novel method for an automatic recognition of spike trains in voltage traces, which eliminates the need for human intervention. This enables classification of waveforms with consistent criteria across all the analyzed traces and so it amounts to reduction of the noise in the data. This method allows for an automatic extraction of relevant physiological parameters necessary for further statistical analysis. In order to illustrate the usefulness of this procedure to analyze voltage traces, we characterized the influence of the somatic current injection level on several electrophysiological parameters in a set of modeled neurons. This application suggests that such an algorithmic processing of physiological data extracts parameters in a suitable form for further investigation of structure-activity relationship in single neurons.

ATD: a multiplatform for semiautomatic 3-D detection of kidneys and their pathology in real time

This research presents a novel multi-functional system for medical Imaging-enabled Assistive Diagnosis (IAD). Although the IAD demonstrator has focused on abdominal images and supports the clinical diagnosis of kidneys using CT/MRI imaging, it can be adapted to work on image delineation, annotation and 3D real-size volumetric modelling of other organ structures such as the brain, spine, etc. The IAD provides advanced real-time 3D visualisation and measurements with fully automated functionalities as developed in two stages. In the first stage, via the clinically driven user interface, specialist clinicians use CT/MRI imaging datasets to accurately delineate and annotate the kidneys and their possible abnormalities, thus creating “3D Golden Standard Models”. Based on these models, in the second stage, clinical support staff i.e. medical technicians interactively define model-based rules and parameters for the integrated “Automatic Recognition Framework” to achieve results which are closest to that of the clinicians. These specific rules and parameters are stored in “Templates” and can later be used by any clinician to automatically identify organ structures i.e. kidneys and their possible abnormalities. The system also supports the transmission of these “Templates” to another expert for a second opinion. A 3D model of the body, the organs and their possible pathology with real metrics is also integrated. The automatic functionality was tested on eleven MRI datasets (comprising of 286 images) and the 3D models were validated by comparing them with the metrics from the corresponding “3D Golden Standard Models”. The system provides metrics for the evaluation of the results, in terms of Accuracy, Precision, Sensitivity, Specificity and Dice Similarity Coefficient (DSC) so as to enable benchmarking of its performance. The first IAD prototype has produced promising results as its performance accuracy based on the most widely deployed evaluation metric, DSC, yields 97% for the recognition of kidneys and 96% for their abnormalities; whilst across all the above evaluation metrics its performance ranges between 96% and 100%. Further development of the IAD system is in progress to extend and evaluate its clinical diagnostic support capability through development and integration of additional algorithms to offer fully computer-aided identification of other organs and their abnormalities based on CT/MRI/Ultra-sound Imaging.

Optimizing optimum-path forest classification for huge datasets

Traditional pattern recognition techniques can not handle the classification of large datasets with both efficiency and effectiveness. In this context, the Optimum-Path Forest (OPF) classifier was recently introduced, trying to achieve high recognition rates and low computational cost. Although OPF was much faster than Support Vector Machines for training, it was slightly slower for classification. In this paper, we present the Efficient OPF (EOPF), which is an enhanced and faster version of the traditional OPF, and validate it for the automatic recognition of white matter and gray matter in magnetic resonance images of the human brain. © 2010 IEEE.

Parkinson's disease identification through Optimum-Path Forest

Artificial intelligence techniques have been extensively used for the identification of several disorders related with the voice signal analysis, such as Parkinson's disease (PD). However, some of these techniques flaw by assuming some separability in the original feature space or even so in the one induced by a kernel mapping. In this paper we propose the PD automatic recognition by means of Optimum-Path Forest (OPF), which is a new recently developed pattern recognition technique that does not assume any shape/separability of the classes/feature space. The experiments showed that OPF outperformed Support Vector Machines, Artificial Neural Networks and other commonly used supervised classification techniques for PD identification. © 2010 IEEE.

Estudo de métodos para classificação e localização precisa de padrões usando um sistema de luz estruturada

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Análise comparativa dos resultados obtidos pelos métodos de calibração de campo tridimensional e bidimensional

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

Insolvenza transfrontaliera e giurisdizione: l'esperienza del regolamento (CE) n. 1346 del 2000

L’elaborato è strutturato in quattro parti: la prima è dedicata all’inquadramento del background normativo. L’autrice affronta, con un approccio ricostruttivo, i precedenti alla redazione del regolamento e le difficoltà incontrate a causa delle resistenze degli Stati membri. Si sofferma altresì sulle norme UNCITRAL, anch’esse oggetto nel recente periodo, di numerose implementazioni. La seconda parte fotografa il ruolo della Corte di Giustizia nell’interpretazione del regolamento n. 1346/2000 ed individua i concetti fondamentali del regolamento: l’universalità attenuata, il campo di applicazione soggettivo del regolamento, la legge applicabile, il principio dell’automatico riconoscimento delle decisioni e la correlata tematica dell’ordine pubblico, nonché la figura del curatore. Si approfondisce l’attività degli Stati membri nel dotarsi di norme di coordinamento ( esemplare il caso della Spagna e della Germania) sottolineando il silenzio del legislatore italiano che, nonostante le numerose riforme in materia, a tutt’oggi non ha ideato un sistema in grado di coordinare la normativa nazionale con la struttura del regolamento europeo. Nella terza parte l'autrice approfondisce la giurisdizione nel regolamento n. 1346/2000. Si individuano le parole chiave: Comi e dipendenza, i cui significati sono sfumati seguendo le posizioni della Corte di Giustizia, (Leading Case Eurofood sino a Interedil) e si mette in discussione, nel panorama attuale, la tenuta di tali criteri giurisdizionali. Sempre intorno al concetto di Comi, si analizzano: la giurisdizione verso gruppi di imprese disciplina assente nel regolamento, i rapporti tra procedura principale e secondaria , la giurisdizione in materia di azioni connesse e/o correlate (Gourdain vs Nadler/Seagon vs Deko Marty). Il capitolo conclusivo offre una panoramica delle proposte finalizzate ad un’implementazione della struttura del regolamento sull’insolvenza. Numerose, infatti, sono le proposte a livello dottrinale e da parte degli organi comunitari in vista della scadenza del Report della Commissione Europea sulla applicazione del regolamento 1346 del 2000 (art. 46).

Risk analysis of tractor rollover in the normal operation in field

Tractor rollover represent a primary cause of death or serious injury in agriculture and despite the mandatory Roll-Over Protective Structures (ROPS), that reduced the number of injuries, tractor accidents are still of great concern. Because of their versatility and wide use many studies on safety are concerned with the stability of tractors, but they often prefer controlled tests or laboratory tests. The evaluation of tractors working in field, instead, is a very complex issue because the rollover could be influenced by the interaction among operator, tractor and environment. Recent studies are oriented towards the evaluation of the actual working conditions developing prototypes for driver assistance and data acquisition. Currently these devices are produced and sold by manufacturers. A warning device was assessed in this study with the aim to evaluate its performance and to collect data on different variables influencing the dynamics of tractors in field by monitoring continuously the working conditions of tractors operating at the experimental farm of the Bologna University. The device consists of accelerometers, gyroscope, GSM/GPRS, GPS for geo-referencing and a transceiver for the automatic recognition of tractor-connected equipment. A microprocessor processes data and provides information, through a dedicated algorithm requiring data on the geometry of the tested tractor, on the level of risk for the operator in terms of probable loss of stability and suggests corrective measures to reduce the potential instability of the tractor.

Structural patterns for document engineering: from an empirical bottom-up analysis to an ontological theory

This thesis aims at investigating a new approach to document analysis based on the idea of structural patterns in XML vocabularies. My work is founded on the belief that authors do naturally converge to a reasonable use of markup languages and that extreme, yet valid instances are rare and limited. Actual documents, therefore, may be used to derive classes of elements (patterns) persisting across documents and distilling the conceptualization of the documents and their components, and may give ground for automatic tools and services that rely on no background information (such as schemas) at all. The central part of my work consists in introducing from the ground up a formal theory of eight structural patterns (with three sub-patterns) that are able to express the logical organization of any XML document, and verifying their identifiability in a number of different vocabularies. This model is characterized by and validated against three main dimensions: terseness (i.e. the ability to represent the structure of a document with a small number of objects and composition rules), coverage (i.e. the ability to capture any possible situation in any document) and expressiveness (i.e. the ability to make explicit the semantics of structures, relations and dependencies). An algorithm for the automatic recognition of structural patterns is then presented, together with an evaluation of the results of a test performed on a set of more than 1100 documents from eight very different vocabularies. This language-independent analysis confirms the ability of patterns to capture and summarize the guidelines used by the authors in their everyday practice. Finally, I present some systems that work directly on the pattern-based representation of documents. The ability of these tools to cover very different situations and contexts confirms the effectiveness of the model.

Seizure detection algorithm for neonates based on wave-sequence analysis

Objective: The description and evaluation of the performance of a new real-time seizure detection algorithm in the newborn infant. Methods: The algorithm includes parallel fragmentation of EEG signal into waves; wave-feature extraction and averaging; elementary, preliminary and final detection. The algorithm detects EEG waves with heightened regularity, using wave intervals, amplitudes and shapes. The performance of the algorithm was assessed with the use of event-based and liberal and conservative time-based approaches and compared with the performance of Gotman's and Liu's algorithms. Results: The algorithm was assessed on multi-channel EEG records of 55 neonates including 17 with seizures. The algorithm showed sensitivities ranging 83-95% with positive predictive values (PPV) 48-77%. There were 2.0 false positive detections per hour. In comparison, Gotman's algorithm (with 30 s gap-closing procedure) displayed sensitivities of 45-88% and PPV 29-56%; with 7.4 false positives per hour and Liu's algorithm displayed sensitivities of 96-99%, and PPV 10-25%; with 15.7 false positives per hour. Conclusions: The wave-sequence analysis based algorithm displayed higher sensitivity, higher PPV and a substantially lower level of false positives than two previously published algorithms. Significance: The proposed algorithm provides a basis for major improvements in neonatal seizure detection and monitoring. Published by Elsevier Ireland Ltd. on behalf of International Federation of Clinical Neurophysiology.

Computational methods for the detection of facial palsy

We are developing a telemedicine application which offers automated diagnosis of facial (Bell's) palsy through a Web service. We used a test data set of 43 images of facial palsy patients and 44 normal people to develop the automatic recognition algorithm. Three different image pre-processing methods were used. Machine learning techniques (support vector machine, SVM) were used to examine the difference between the two halves of the face. If there was a sufficient difference, then the SVM recognized facial palsy. Otherwise, if the halves were roughly symmetrical, the SVM classified the image as normal. It was found that the facial palsy images had a greater Hamming Distance than the normal images, indicating greater asymmetry. The median distance in the normal group was 331 (interquartile range 277-435) and the median distance in the facial palsy group was 509 (interquartile range 334-703). This difference was significant (P