A framework for developing motion-based games

Dissertação para obtenção do Grau de Mestre em Engenharia Informática

CMOS indoor light energy harvesting system for wireless sensing applications

Dissertação para obtenção do Grau de Doutor em Engenharia Electrotécnica e de Computadores

2D position system for a mobile robot in unstructured environments

Nowadays, several sensors and mechanisms are available to estimate a mobile robot trajectory and location with respect to its surroundings. Usually absolute positioning mechanisms are the most accurate, but they also are the most expensive ones, and require pre installed equipment in the environment. Therefore, a system capable of measuring its motion and location within the environment (relative positioning) has been a research goal since the beginning of autonomous vehicles. With the increasing of the computational performance, computer vision has become faster and, therefore, became possible to incorporate it in a mobile robot. In visual odometry feature based approaches, the model estimation requires absence of feature association outliers for an accurate motion. Outliers rejection is a delicate process considering there is always a trade-off between speed and reliability of the system. This dissertation proposes an indoor 2D position system using Visual Odometry. The mobile robot has a camera pointed to the ceiling, for image analysis. As requirements, the ceiling and the oor (where the robot moves) must be planes. In the literature, RANSAC is a widely used method for outlier rejection. However, it might be slow in critical circumstances. Therefore, it is proposed a new algorithm that accelerates RANSAC, maintaining its reliability. The algorithm, called FMBF, consists on comparing image texture patterns between pictures, preserving the most similar ones. There are several types of comparisons, with different computational cost and reliability. FMBF manages those comparisons in order to optimize the trade-off between speed and reliability.

Enrollment time as a requirement for biometric hand recognition systems

Biometric systems are increasingly being used as a means for authentication to provide system security in modern technologies. The performance of a biometric system depends on the accuracy, the processing speed, the template size, and the time necessary for enrollment. While much research has focused on the first three factors, enrollment time has not received as much attention. In this work, we present the findings of our research focused upon studying user’s behavior when enrolling in a biometric system. Specifically, we collected information about the user’s availability for enrollment in respect to the hand recognition systems (e.g., hand geometry, palm geometry or any other requiring positioning the hand on an optical scanner). A sample of 19 participants, chosen randomly apart their age, gender, profession and nationality, were used as test subjects in an experiment to study the patience of users enrolling in a biometric hand recognition system.

Sparse nonlinear discriminants

Kernel-Functions, Machine Learning, Least Squares, Speech Recognition, Classification, Regression

A New Large-DNA-Fragment Delivery System Based on Integrase Activity from an Integrative and Conjugative Element.

During the past few decades, numerous plasmid vectors have been developed for cloning, gene expression analysis, and genetic engineering. Cloning procedures typically rely on PCR amplification, DNA fragment restriction digestion, recovery, and ligation, but increasingly, procedures are being developed to assemble large synthetic DNAs. In this study, we developed a new gene delivery system using the integrase activity of an integrative and conjugative element (ICE). The advantage of the integrase-based delivery is that it can stably introduce a large DNA fragment (at least 75 kb) into one or more specific sites (the gene for glycine-accepting tRNA) on a target chromosome. Integrase recombination activity in Escherichia coli is kept low by using a synthetic hybrid promoter, which, however, is unleashed in the final target host, forcing the integration of the construct. Upon integration, the system is again silenced. Two variants with different genetic features were produced, one in the form of a cloning vector in E. coli and the other as a mini-transposable element by which large DNA constructs assembled in E. coli can be tagged with the integrase gene. We confirmed that the system could successfully introduce cosmid and bacterial artificial chromosome (BAC) DNAs from E. coli into the chromosome of Pseudomonas putida in a site-specific manner. The integrase delivery system works in concert with existing vector systems and could thus be a powerful tool for synthetic constructions of new metabolic pathways in a variety of host bacteria.

Object recognition for autonomous objects: comparison of two approaches

Report for the scientific sojourn at the Swiss Federal Institute of Technology Zurich, Switzerland, between September and December 2007. In order to make robots useful assistants for our everyday life, the ability to learn and recognize objects is of essential importance. However, object recognition in real scenes is one of the most challenging problems in computer vision, as it is necessary to deal with difficulties. Furthermore, in mobile robotics a new challenge is added to the list: computational complexity. In a dynamic world, information about the objects in the scene can become obsolete before it is ready to be used if the detection algorithm is not fast enough. Two recent object recognition techniques have achieved notable results: the constellation approach proposed by Lowe and the bag of words approach proposed by Nistér and Stewénius. The Lowe constellation approach is the one currently being used in the robot localization project of the COGNIRON project. This report is divided in two main sections. The first section is devoted to briefly review the currently used object recognition system, the Lowe approach, and bring to light the drawbacks found for object recognition in the context of indoor mobile robot navigation. Additionally the proposed improvements for the algorithm are described. In the second section the alternative bag of words method is reviewed, as well as several experiments conducted to evaluate its performance with our own object databases. Furthermore, some modifications to the original algorithm to make it suitable for object detection in unsegmented images are proposed.

Flow cytometry as a tool to identify Mycobacterium tuberculosis interaction with the immune system and drug susceptibility

Flow cytometric analysis is a useful and widely employed tool to identify immunological alterations caused by different microorganisms, including Mycobacterium tuberculosis. However, this tool can be used for several others analysis. We will discuss some applications for flow cytometry to the study of M. tuberculosis, mainly on cell surface antigens, mycobacterial secreted proteins, their interaction with the immune system using inflammatory cells recovered from peripheral blood, alveolar and pleura spaces and the influence of M. tuberculosis on apoptosis, and finally the rapid determination of drug susceptibility. All of these examples highlight the usefulness of flow cytometry in the study of M. tuber-culosis infection.

Intégration de méthodes de data mining dans le renseignement criminel : analyse par des structures issues de la théorie des graphes dans le profilage des stupéfiants

Data mining can be defined as the extraction of previously unknown and potentially useful information from large datasets. The main principle is to devise computer programs that run through databases and automatically seek deterministic patterns. It is applied in different fields of application, e.g., remote sensing, biometry, speech recognition, but has seldom been applied to forensic case data. The intrinsic difficulty related to the use of such data lies in its heterogeneity, which comes from the many different sources of information. The aim of this study is to highlight potential uses of pattern recognition that would provide relevant results from a criminal intelligence point of view. The role of data mining within a global crime analysis methodology is to detect all types of structures in a dataset. Once filtered and interpreted, those structures can point to previously unseen criminal activities. The interpretation of patterns for intelligence purposes is the final stage of the process. It allows the researcher to validate the whole methodology and to refine each step if necessary. An application to cutting agents found in illicit drug seizures was performed. A combinatorial approach was done, using the presence and the absence of products. Methods coming from the graph theory field were used to extract patterns in data constituted by links between products and place and date of seizure. A data mining process completed using graphing techniques is called ``graph mining''. Patterns were detected that had to be interpreted and compared with preliminary knowledge to establish their relevancy. The illicit drug profiling process is actually an intelligence process that uses preliminary illicit drug classes to classify new samples. Methods proposed in this study could be used \textit{a priori} to compare structures from preliminary and post-detection patterns. This new knowledge of a repeated structure may provide valuable complementary information to profiling and become a source of intelligence.

A wearable system for multi-segment foot kinetics measurement.

This study aims to design a wearable system for kinetics measurement of multi-segment foot joints in long-distance walking and to investigate its suitability for clinical evaluations. The wearable system consisted of inertial sensors (3D gyroscopes and 3D accelerometers) on toes, forefoot, hindfoot, and shank, and a plantar pressure insole. After calibration in a laboratory, 10 healthy elderly subjects and 12 patients with ankle osteoarthritis walked 50m twice wearing this system. Using inverse dynamics, 3D forces, moments, and power were calculated in the joint sections among toes, forefoot, hindfoot, and shank. Compared to those we previously estimated for a one-segment foot model, the sagittal and transverse moments and power in the ankle joint, as measured via multi-segment foot model, showed a normalized RMS difference of less than 11%, 14%, and 13%, respectively, for healthy subjects, and 13%, 15%, and 14%, for patients. Similar to our previous study, the coronal moments were not analyzed. Maxima-minima values of anterior-posterior and vertical force, sagittal moment, and power in shank-hindfoot and hindfoot-forefoot joints were significantly different between patients and healthy subjects. Except for power, the inter-subject repeatability of these parameters was CMC>0.90 for healthy subjects and CMC>0.70 for patients. Repeatability of these parameters was lower for the forefoot-toes joint. The proposed measurement system estimated multi-segment foot joints kinetics with acceptable repeatability but showed difference, compared to those previously estimated for the one-segment foot model. These parameters also could distinguish patients from healthy subjects. Thus, this system is suggested for outcome evaluations of foot treatments.

Reconeixement de veu mitjançant xarxes neuronals

La interacció home-màquina per mitjà de la veu cobreix moltes àrees d’investigació. Es destaquen entre altres, el reconeixement de la parla, la síntesis i identificació de discurs, la verificació i identificació de locutor i l’activació per veu (ordres) de sistemes robòtics. Reconèixer la parla és natural i simple per a les persones, però és un treball complex per a les màquines, pel qual existeixen diverses metodologies i tècniques, entre elles les Xarxes Neuronals. L’objectiu d’aquest treball és desenvolupar una eina en Matlab per al reconeixement i identificació de paraules pronunciades per un locutor, entre un conjunt de paraules possibles, i amb una bona fiabilitat dins d’uns marges preestablerts. El sistema és independent del locutor que pronuncia la paraula, és a dir, aquest locutor no haurà intervingut en el procés d’entrenament del sistema. S’ha dissenyat una interfície que permet l’adquisició del senyal de veu i el seu processament mitjançant xarxes neuronals i altres tècniques. Adaptant una part de control al sistema, es podria utilitzar per donar ordres a un robot com l’Alfa6Uvic o qualsevol altre dispositiu.

Reconeixement de gestos de la mà amb el sensor Kinect

El reconeixement dels gestos de la mà (HGR, Hand Gesture Recognition) és actualment un camp important de recerca degut a la varietat de situacions en les quals és necessari comunicar-se mitjançant signes, com pot ser la comunicació entre persones que utilitzen la llengua de signes i les que no. En aquest projecte es presenta un mètode de reconeixement de gestos de la mà a temps real utilitzant el sensor Kinect per Microsoft Xbox, implementat en un entorn Linux (Ubuntu) amb llenguatge de programació Python i utilitzant la llibreria de visió artifical OpenCV per a processar les dades sobre un ordinador portàtil convencional. Gràcies a la capacitat del sensor Kinect de capturar dades de profunditat d’una escena es poden determinar les posicions i trajectòries dels objectes en 3 dimensions, el que implica poder realitzar una anàlisi complerta a temps real d’una imatge o d’una seqüencia d’imatges. El procediment de reconeixement que es planteja es basa en la segmentació de la imatge per poder treballar únicament amb la mà, en la detecció dels contorns, per després obtenir l’envolupant convexa i els defectes convexos, que finalment han de servir per determinar el nombre de dits i concloure en la interpretació del gest; el resultat final és la transcripció del seu significat en una finestra que serveix d’interfície amb l’interlocutor. L’aplicació permet reconèixer els números del 0 al 5, ja que s’analitza únicament una mà, alguns gestos populars i algunes de les lletres de l’alfabet dactilològic de la llengua de signes catalana. El projecte és doncs, la porta d’entrada al camp del reconeixement de gestos i la base d’un futur sistema de reconeixement de la llengua de signes capaç de transcriure tant els signes dinàmics com l’alfabet dactilològic.

MM-GBSA binding free energy decomposition and T cell receptor engineering.

Recognition by the T-cell receptor (TCR) of immunogenic peptides (p) presented by class I major histocompatibility complexes (MHC) is the key event in the immune response against virus infected cells or tumor cells. The major determinant of T cell activation is the affinity of the TCR for the peptide-MHC complex, though kinetic parameters are also important. A study of the 2C TCR/SIYR/H-2Kb system using a binding free energy decomposition (BFED) based on the MM-GBSA approach had been performed to assess the performance of the approach on this system. The results showed that the TCR-p-MHC BFED including entropic terms provides a detailed and reliable description of the energetics of the interaction (Zoete and Michielin, 2007). Based on these results, we have developed a new approach to design sequence modifications for a TCR recognizing the human leukocyte antigen (HLA)-A2 restricted tumor epitope NY-ESO-1. NY-ESO-1 is a cancer testis antigen expressed not only in melanoma, but also on several other types of cancers. It has been observed at high frequencies in melanoma patients with unusually positive clinical outcome and, therefore, represents an interesting target for adoptive transfer with modified TCR. Sequence modifications of TCR potentially increasing the affinity for this epitope have been proposed and tested in vitro. T cells expressing some of the proposed TCR mutants showed better T cell functionality, with improved killing of peptide-loaded T2 cells and better proliferative capacity compared to the wild type TCR expressing cells. These results open the door of rational TCR design for adoptive transfer cancer therapy.

MRI of coronary vessel walls using radial k-space sampling and steady-state free precession imaging.

OBJECTIVE: The objective of our study was to investigate the impact of radial k-space sampling and steady-state free precession (SSFP) imaging on image quality in MRI of coronary vessel walls. SUBJECTS AND METHODS: Eleven subjects were examined on a 1.5-T MR system using three high-resolution navigator-gated and cardiac-triggered 3D black blood sequences (cartesian gradient-echo [GRE], radial GRE, and radial SSFP) with identical spatial resolution (0.9 x 0.9 x 2.4 mm3). The signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), vessel wall sharpness, and motion artifacts were analyzed. RESULTS: The mean SNR and CNR of the coronary vessel wall were improved using radial imaging and were best using radial k-space sampling combined with SSFP imaging. Vessel border definition was similar for all three sequences. Radial k-space sampling was found to be less sensitive to motion. Consistently good image quality was seen with the radial GRE sequence. CONCLUSION: Radial k-space sampling in MRI of coronary vessel walls resulted in fewer motion artifacts and improved SNR and CNR. The use of SSFP imaging, however, did not result in improved coronary vessel wall visualization.