MIRACLE’s Ad-Hoc and Geographical IR Approaches for CLEF 2006

This paper presents the 2006 Miracle team’s approaches to the Ad-Hoc and Geographical Information Retrieval tasks. A first set of runs was obtained using a set of basic components. Then, by putting together special combinations of these runs, an extended set was obtained. With respect to previous campaigns some improvements have been introduced in our system: an entity recognition prototype is integrated in our tokenization scheme, and the performance of our indexing and retrieval engine has been improved. For GeoCLEF, we tested retrieving using geo-entity and textual references separately, and then combining them with different approaches.

Automatic information and safety systems for driving assistance

O objeto principal desta tese é o estudo de algoritmos de processamento e representação automáticos de dados, em particular de informação obtida por sensores montados a bordo de veículos (2D e 3D), com aplicação em contexto de sistemas de apoio à condução. O trabalho foca alguns dos problemas que, quer os sistemas de condução automática (AD), quer os sistemas avançados de apoio à condução (ADAS), enfrentam hoje em dia. O documento é composto por duas partes. A primeira descreve o projeto, construção e desenvolvimento de três protótipos robóticos, incluindo pormenores associados aos sensores montados a bordo dos robôs, algoritmos e arquitecturas de software. Estes robôs foram utilizados como plataformas de ensaios para testar e validar as técnicas propostas. Para além disso, participaram em várias competições de condução autónoma tendo obtido muito bons resultados. A segunda parte deste documento apresenta vários algoritmos empregues na geração de representações intermédias de dados sensoriais. Estes podem ser utilizados para melhorar técnicas já existentes de reconhecimento de padrões, deteção ou navegação, e por este meio contribuir para futuras aplicações no âmbito dos AD ou ADAS. Dado que os veículos autónomos contêm uma grande quantidade de sensores de diferentes naturezas, representações intermédias são particularmente adequadas, pois podem lidar com problemas relacionados com as diversas naturezas dos dados (2D, 3D, fotométrica, etc.), com o carácter assíncrono dos dados (multiplos sensores a enviar dados a diferentes frequências), ou com o alinhamento dos dados (problemas de calibração, diferentes sensores a disponibilizar diferentes medições para um mesmo objeto). Neste âmbito, são propostas novas técnicas para a computação de uma representação multi-câmara multi-modal de transformação de perspectiva inversa, para a execução de correcção de côr entre imagens de forma a obter mosaicos de qualidade, ou para a geração de uma representação de cena baseada em primitivas poligonais, capaz de lidar com grandes quantidades de dados 3D e 2D, tendo inclusivamente a capacidade de refinar a representação à medida que novos dados sensoriais são recebidos.

Design and control of multi-finger haptic devices for dexterous manipulation

En la interacción con el entorno que nos rodea durante nuestra vida diaria (utilizar un cepillo de dientes, abrir puertas, utilizar el teléfono móvil, etc.) y en situaciones profesionales (intervenciones médicas, procesos de producción, etc.), típicamente realizamos manipulaciones avanzadas que incluyen la utilización de los dedos de ambas manos. De esta forma el desarrollo de métodos de interacción háptica multi-dedo dan lugar a interfaces hombre-máquina más naturales y realistas. No obstante, la mayoría de interfaces hápticas disponibles en el mercado están basadas en interacciones con un solo punto de contacto; esto puede ser suficiente para la exploración o palpación del entorno pero no permite la realización de tareas más avanzadas como agarres. En esta tesis, se investiga el diseño mecánico, control y aplicaciones de dispositivos hápticos modulares con capacidad de reflexión de fuerzas en los dedos índice, corazón y pulgar del usuario. El diseño mecánico de la interfaz diseñada, ha sido optimizado con funciones multi-objetivo para conseguir una baja inercia, un amplio espacio de trabajo, alta manipulabilidad y reflexión de fuerzas superiores a 3 N en el espacio de trabajo. El ancho de banda y la rigidez del dispositivo se han evaluado mediante simulación y experimentación real. Una de las áreas más importantes en el diseño de estos dispositivos es el efector final, ya que es la parte que está en contacto con el usuario. Durante este trabajo se ha diseñado un dedal de bajo peso, adaptable a diferentes usuarios que, mediante la incorporación de sensores de contacto, permite estimar fuerzas normales y tangenciales durante la interacción con entornos reales y virtuales. Para el diseño de la arquitectura de control, se estudiaron los principales requisitos para estos dispositivos. Entre estos, cabe destacar la adquisición, procesado e intercambio a través de internet de numerosas señales de control e instrumentación; la computación de equaciones matemáticas incluyendo la cinemática directa e inversa, jacobiana, algoritmos de detección de agarres, etc. Todos estos componentes deben calcularse en tiempo real garantizando una frecuencia mínima de 1 KHz. Además, se describen sistemas para manipulación de precisión virtual y remota; así como el diseño de un método denominado "desacoplo cinemático iterativo" para computar la cinemática inversa de robots y la comparación con otros métodos actuales. Para entender la importancia de la interacción multimodal, se ha llevado a cabo un estudio para comprobar qué estímulos sensoriales se correlacionan con tiempos de respuesta más rápidos y de mayor precisión. Estos experimentos se desarrollaron en colaboración con neurocientíficos del instituto Technion Israel Institute of Technology. Comparando los tiempos de respuesta en la interacción unimodal (auditiva, visual y háptica) con combinaciones bimodales y trimodales de los mismos, se demuestra que el movimiento sincronizado de los dedos para generar respuestas de agarre se basa principalmente en la percepción háptica. La ventaja en el tiempo de procesamiento de los estímulos hápticos, sugiere que los entornos virtuales que incluyen esta componente sensorial generan mejores contingencias motoras y mejoran la credibilidad de los eventos. Se concluye que, los sistemas que incluyen percepción háptica dotan a los usuarios de más tiempo en las etapas cognitivas para rellenar información de forma creativa y formar una experiencia más rica. Una aplicación interesante de los dispositivos hápticos es el diseño de nuevos simuladores que permitan entrenar habilidades manuales en el sector médico. En colaboración con fisioterapeutas de Griffith University en Australia, se desarrolló un simulador que permite realizar ejercicios de rehabilitación de la mano. Las propiedades de rigidez no lineales de la articulación metacarpofalange del dedo índice se estimaron mediante la utilización del efector final diseñado. Estos parámetros, se han implementado en un escenario que simula el comportamiento de la mano humana y que permite la interacción háptica a través de esta interfaz. Las aplicaciones potenciales de este simulador están relacionadas con entrenamiento y educación de estudiantes de fisioterapia. En esta tesis, se han desarrollado nuevos métodos que permiten el control simultáneo de robots y manos robóticas en la interacción con entornos reales. El espacio de trabajo alcanzable por el dispositivo háptico, se extiende mediante el cambio de modo de control automático entre posición y velocidad. Además, estos métodos permiten reconocer el gesto del usuario durante las primeras etapas de aproximación al objeto para su agarre. Mediante experimentos de manipulación avanzada de objetos con un manipulador y diferentes manos robóticas, se muestra que el tiempo en realizar una tarea se reduce y que el sistema permite la realización de la tarea con precisión. Este trabajo, es el resultado de una colaboración con investigadores de Harvard BioRobotics Laboratory. ABSTRACT When we interact with the environment in our daily life (using a toothbrush, opening doors, using cell-phones, etc.), or in professional situations (medical interventions, manufacturing processes, etc.) we typically perform dexterous manipulations that involve multiple fingers and palm for both hands. Therefore, multi-Finger haptic methods can provide a realistic and natural human-machine interface to enhance immersion when interacting with simulated or remote environments. Most commercial devices allow haptic interaction with only one contact point, which may be sufficient for some exploration or palpation tasks but are not enough to perform advanced object manipulations such as grasping. In this thesis, I investigate the mechanical design, control and applications of a modular haptic device that can provide force feedback to the index, thumb and middle fingers of the user. The designed mechanical device is optimized with a multi-objective design function to achieve a low inertia, a large workspace, manipulability, and force-feedback of up to 3 N within the workspace; the bandwidth and rigidity for the device is assessed through simulation and real experimentation. One of the most important areas when designing haptic devices is the end-effector, since it is in contact with the user. In this thesis the design and evaluation of a thimble-like, lightweight, user-adaptable, and cost-effective device that incorporates four contact force sensors is described. This design allows estimation of the forces applied by a user during manipulation of virtual and real objects. The design of a real-time, modular control architecture for multi-finger haptic interaction is described. Requirements for control of multi-finger haptic devices are explored. Moreover, a large number of signals have to be acquired, processed, sent over the network and mathematical computations such as device direct and inverse kinematics, jacobian, grasp detection algorithms, etc. have to be calculated in Real Time to assure the required high fidelity for the haptic interaction. The Hardware control architecture has different modules and consists of an FPGA for the low-level controller and a RT controller for managing all the complex calculations (jacobian, kinematics, etc.); this provides a compact and scalable solution for the required high computation capabilities assuring a correct frequency rate for the control loop of 1 kHz. A set-up for dexterous virtual and real manipulation is described. Moreover, a new algorithm named the iterative kinematic decoupling method was implemented to solve the inverse kinematics of a robotic manipulator. In order to understand the importance of multi-modal interaction including haptics, a subject study was carried out to look for sensory stimuli that correlate with fast response time and enhanced accuracy. This experiment was carried out in collaboration with neuro-scientists from Technion Israel Institute of Technology. By comparing the grasping response times in unimodal (auditory, visual, and haptic) events with the response times in events with bimodal and trimodal combinations. It is concluded that in grasping tasks the synchronized motion of the fingers to generate the grasping response relies on haptic cues. This processing-speed advantage of haptic cues suggests that multimodalhaptic virtual environments are superior in generating motor contingencies, enhancing the plausibility of events. Applications that include haptics provide users with more time at the cognitive stages to fill in missing information creatively and form a richer experience. A major application of haptic devices is the design of new simulators to train manual skills for the medical sector. In collaboration with physical therapists from Griffith University in Australia, we developed a simulator to allow hand rehabilitation manipulations. First, the non-linear stiffness properties of the metacarpophalangeal joint of the index finger were estimated by using the designed end-effector; these parameters are implemented in a scenario that simulates the behavior of the human hand and that allows haptic interaction through the designed haptic device. The potential application of this work is related to educational and medical training purposes. In this thesis, new methods to simultaneously control the position and orientation of a robotic manipulator and the grasp of a robotic hand when interacting with large real environments are studied. The reachable workspace is extended by automatically switching between rate and position control modes. Moreover, the human hand gesture is recognized by reading the relative movements of the index, thumb and middle fingers of the user during the early stages of the approximation-to-the-object phase and then mapped to the robotic hand actuators. These methods are validated to perform dexterous manipulation of objects with a robotic manipulator, and different robotic hands. This work is the result of a research collaboration with researchers from the Harvard BioRobotics Laboratory. The developed experiments show that the overall task time is reduced and that the developed methods allow for full dexterity and correct completion of dexterous manipulations.

Signal processing for molecular and cellular biological physics:an emerging field

Recent advances in our ability to watch the molecular and cellular processes of life in action-such as atomic force microscopy, optical tweezers and Forster fluorescence resonance energy transfer-raise challenges for digital signal processing (DSP) of the resulting experimental data. This article explores the unique properties of such biophysical time series that set them apart from other signals, such as the prevalence of abrupt jumps and steps, multi-modal distributions and autocorrelated noise. It exposes the problems with classical linear DSP algorithms applied to this kind of data, and describes new nonlinear and non-Gaussian algorithms that are able to extract information that is of direct relevance to biological physicists. It is argued that these new methods applied in this context typify the nascent field of biophysical DSP. Practical experimental examples are supplied.

2D-3D Face Recognition Based on PCA and Feature Modelling

Hybrid face recognition, using image (2D) and structural (3D) information, has explored the fusion of Nearest Neighbour classifiers. This paper examines the effectiveness of feature modelling for each individual modality, 2D and 3D. Furthermore, it is demonstrated that the fusion of feature modelling techniques for the 2D and 3D modalities yields performance improvements over the individual classifiers. By fusing the feature modelling classifiers for each modality with equal weights the average Equal Error Rate improves from 12.60% for the 2D classifier and 12.10% for the 3D classifier to 7.38% for the Hybrid 2D+3D clasiffier.

Sequential decision fusion for controlled detection errors

Information fusion in biometrics has received considerable attention. The architecture proposed here is based on the sequential integration of multi-instance and multi-sample fusion schemes. This method is analytically shown to improve the performance and allow a controlled trade-off between false alarms and false rejects when the classifier decisions are statistically independent. Equations developed for detection error rates are experimentally evaluated by considering the proposed architecture for text dependent speaker verification using HMM based digit dependent speaker models. The tuning of parameters, n classifiers and m attempts/samples, is investigated and the resultant detection error trade-off performance is evaluated on individual digits. Results show that performance improvement can be achieved even for weaker classifiers (FRR-19.6%, FAR-16.7%). The architectures investigated apply to speaker verification from spoken digit strings such as credit card numbers in telephone or VOIP or internet based applications.

Attribute exclusion strategies in airline choice : accounting for exogenous information on decision maker processing strategies in models of discrete choice

We examine the impact of individual-specific information processing strategies (IPSs) on the inclusion/exclusion of attributes on the parameter estimates and behavioural outputs of models of discrete choice. Current practice assumes that individuals employ a homogenous IPS with regards to how they process attributes of stated choice (SC) experiments. We show how information collected exogenous of the SC experiment on whether respondents either ignored or considered each attribute may be used in the estimation process, and how such information provides outputs that are IPS segment specific. We contend that accounting the inclusion/exclusion of attributes will result in behaviourally richer population parameter estimates.

Visual voice activity detection using frontal versus profile views

Visual activity detection of lip movements can be used to overcome the poor performance of voice activity detection based solely in the audio domain, particularly in noisy acoustic conditions. However, most of the research conducted in visual voice activity detection (VVAD) has neglected addressing variabilities in the visual domain such as viewpoint variation. In this paper we investigate the effectiveness of the visual information from the speaker’s frontal and profile views (i.e left and right side views) for the task of VVAD. As far as we are aware, our work constitutes the first real attempt to study this problem. We describe our visual front end approach and the Gaussian mixture model (GMM) based VVAD framework, and report the experimental results using the freely available CUAVE database. The experimental results show that VVAD is indeed possible from profile views and we give a quantitative comparison of VVAD based on frontal and profile views The results presented are useful in the development of multi-modal Human Machine Interaction (HMI) using a single camera, where the speaker’s face may not always be frontal.

Individual differences in the processing of punishment and reward cues : an application to road safety messages

This thesis examined the extent to which individual differences, as conceptualised by the revised Reinforcement Sensitivity Theory, influenced young drivers' information processing and subsequent acceptance of anti-speeding messages. Using a multi-method approach, the findings highlighted the utility of combining objective measures (a cognitive response time task and electroencephalography) with self-report measures to assess message processing and message acceptance, respectively. This body of research indicated that responses to anti-speeding messages may differ depending on an individual's personality disposition. Overall, the research provided further insight into the development of message strategies to target high risk drivers.

NEU-MODEL : a multi-level object-oriented dynamic emulation laboratory

Computational neuroscience aims to elucidate the mechanisms of neural information processing and population dynamics, through a methodology of incorporating biological data into complex mathematical models. Existing simulation environments model at a particular level of detail; none allow a multi-level approach to neural modelling. Moreover, most are not engineered to produce compute-efficient solutions, an important issue because sufficient processing power is a major impediment in the field. This project aims to apply modern software engineering techniques to create a flexible high performance neural modelling environment, which will allow rigorous exploration of model parameter effects, and modelling at multiple levels of abstraction.

Process visualization techniques for multi-perspective process comparisons

Organizations executing similar business processes need to understand the differences and similarities in activities performed across work environments. Presently, research interest is directed towards the potential of visualization for the display of process models, to support users in their analysis tasks. Although recent literature in process mining and comparison provide several methods and algorithms to perform process and log comparison, few contributions explore novel visualization approaches. This paper analyses process comparison from a design perspective, providing some practical visualization techniques as anal- ysis solutions (/to support process analysis). The design of the visual comparison has been tackled through three different points of view: the general model, the projected model and the side-by-side comparison in order to support the needs of business analysts. A case study is presented showing the application of process mining and visualization techniques to patient treatment across two Australian hospitals.

How Effective Is The Data On Co-Occurrence Of Domains In Multi-Domain Proteins In Prediction Of Protein-Protein Interactions?

We explore the fuse of information on co-occurrence of domains in multi-domain proteins in predicting protein-protein interactions. The basic premise of our work is the assumption that domains co-occurring in a polypeptide chain undergo either structural or functional interactions among themselves. In this study we use a template dataset of domains in multidomain proteins and predict protein-protein interactions in a target organism. We note that maximum number of correct predictions of interacting protein domain families (158) is made in S. cerevisiae when the dataset of closely related organisms is used as the template followed by the more diverse dataset of bacterial proteins (48) and a dataset of randomly chosen proteins (23). We conclude that use of multi-domain information from organisms closely-related to the target can aid prediction of interacting protein families.

Entanglement distribution with global control in a star-shaped multi-splitter

Distributed quantum information processing (QIP) is a promising way to bypass problems due to unwanted interactions between elements. However, this strategy presupposes the engineering of protocols for remote processors. In many of them, pairwise entanglement is a key resource. We study a model which distributes entanglement among elements of a delocalized network without local control. The model is efficient both in finite- and infinite-dimensional Hilbert spaces. We suggest a setup of electromechanical systems to implement our proposal.

Modal Extraction for Wind Turbines using Moving Window Subspace Identification

This paper proposes a method for wind turbine mode identification using the multivariable output error statespace (MOESP) identification algorithm. The paper incorporates a fast moving window QR decomposition and propagator method from array signal processing, yielding a moving window subspace identification algorithm. The algorithm assumes that the system order is known as a priori and remains constant during identification. For the purpose of extracting modal information for turbines modelled as a linear parameter varying (LPV) system, the algorithm is applicable since a nonlinear system can be approximated as a piecewise time invariant system in consecutive data windows. The algorithm is exemplified using numerical simulations which show that the moving window algorithm can track the modal information. The paper also demonstrates that the low computational burden of the algorithm, compared to conventional batch subspace identification, has significant implications for online implementation.

An Event Driven Fusion Approach for Enjoyment Recognition in Real-time

Social signals and interpretation of carried information is of high importance in Human Computer Interaction. Often used for affect recognition, the cues within these signals are displayed in various modalities. Fusion of multi-modal signals is a natural and interesting way to improve automatic classification of emotions transported in social signals. Throughout most present studies, uni-modal affect recognition as well as multi-modal fusion, decisions are forced for fixed annotation segments across all modalities. In this paper, we investigate the less prevalent approach of event driven fusion, which indirectly accumulates asynchronous events in all modalities for final predictions. We present a fusion approach, handling short-timed events in a vector space, which is of special interest for real-time applications. We compare results of segmentation based uni-modal classification and fusion schemes to the event driven fusion approach. The evaluation is carried out via detection of enjoyment-episodes within the audiovisual Belfast Story-Telling Corpus.