Showing papers by "INESC-ID published in 2003"

Misep—linear and nonlinear ICA based on mutual information

Abstract: Linear Independent Components Analysis (ICA) has become an important signal processing and data analysis technique, the typical application being blind source separation in a wide range of signals, such as biomedical, acoustical and astrophysical ones. Nonlinear ICA is less developed, but has the potential to become at least as powerful.This paper presents MISEP, an ICA technique for linear and nonlinear mixtures, which is based on the minimization of the mutual information of the estimated components. MISEP is a generalization of the popular INFOMAX technique, which is extended in two ways: (1) to deal with nonlinear mixtures, and (2) to be able to adapt to the actual statistical distributions of the sources, by dynamically estimating the nonlinearities to be used at the outputs. The resulting MISEP method optimizes a network with a specialized architecture, with a single objective function: the output entropy.The paper also briefly discusses the issue of nonlinear source separation. Examples of linear and nonlinear source separation performed by MISEP are presented.

Tracking Groups of Pedestrians in Video Sequences

Abstract: This paper describes an algorithm for tracking groups of objects in video sequences. The main difficulties addressed in this work concern total occlusions of the objects to be tracked as well as group merging and splitting. A two layer solution is proposed to overcome these difficulties. The first layer produces a set of spatio temporal strokes based on low level operations which manage to track the active regions most of the time. The second layer performs a consistent labeling of the detected segments using a statistical model based on Bayesian networks. The Bayesian network is recursively computed during the tracking operation and allows the update of the tracker results everytime new information is available. Experimental tests are included to show the performance of the algorithm in ambiguous situations.

RDSP: a RISC DSP based on residue number system

Abstract: This paper is focused on low power programmable fast digital signal processors (DSP) design based on a configurable 5-stage RISC core architecture and on residue number systems (RNS). Several innovative aspects are introduced at the control and datapath architecture levels, which support both the binary system and the RNS. A new moduli set {2/sup n/-1, 2/sup 2n/, 2/sup n/+1} is also proposed for balancing the processing time in the different RNS channels. Experimental results, obtained trough RDSP implementation on FPGA and ASIC, show that not only a significant reduction in circuit area and power consumption but also a speedup may be achieved with RNS when compared with a binary DSP.

An Empirical Comparison of Text Categorization Methods

Abstract: In this paper we present a comprehensive comparison of the performance of a number of text categorization methods in two different data sets. In particular, we evaluate the Vector and Latent Semantic Analysis (LSA) methods, a classifier based on Support Vector Machines (SVM) and the k-Nearest Neighbor variations of the Vector and LSA models.

Tell Me That Bit Again... Bringing Interactivity to a Virtual Storyteller

Abstract: Stories and storytelling are a constant presence in our lives since very early childhood. Who does not remember a story narrated by a good storyteller? However, real human storytellers do not always tell the story the same way. They observe their "audience" and adapt the way they are telling the story to better respond to their reactions. This paper focuses on how to bring interactivity to a virtual storyteller by allow- ing users to influence the story. The storyteller is a synthetic 3D granddad that uses voice, gestures and facial expressions to convey the story content to be told. The character's behaviour and the way the story is narrated, is influenced by the user's input. Such input is done by a tangible interface (a kind of mail box) where children put the cards they want in order to influence what will happen in the story being told. A preliminary usability test was made with sixteen children, with ages between nine and ten years old. The results showed that the way interactivity is introduced was quite successful.

Analog Macromodeling using Kernel Methods

Abstract: In this paper we explore the potential of using a general class of functional representation techniques, kernel-based regression, in the nonlinear model reduction problem. The kernel-based viewpoint provides a convenient computational framework for regression, unifying and extending the previously proposed polynomial and piecewise-linear reduction methods. Furthermore, as many familiar methods for linear system manipulation can be leveraged in a nonlinear context, kernels provide insight into how new, more powerful, nonlinear modeling strategies can be constructed. We present an SVD-like technique for automatic compression of nonlinear models that allows systematic identification of model redundancies and rigorous control of approximation error.

Comparison and Combination of Features in a Hybrid HMM/MLP and a HMM/GMM Speech Recognition System

Abstract: Recently, the advantages of the spectral parameters obtained by frequency filtering (FF) of the logarithmic filter-bank energies (logFBEs) have been reported. These parameters, which are frequency derivatives of the lofFBEs, lie in the frequency domain, and have shown good recognition performance with repect to the conventional MFCCs for HMM systems. In this paper, the FF features are first compared with the MFCCs and the Rasta-PLP features using both a hybrid HMM/MLP and a usual HMM/GMM recognition system, for both clean and noisy speech. Taking advantage of the ability of the hybrid system to deal with correlated features, the inclusion of both the frequency second-derivatives and the raw logFBes as additional features is proposed and tested. Moreover, the robustness of these features in noisy conditions is enhanced by combining the FF technique with the Rasta temporal filtering approach. Finally, a study of the FF features in the framework of multi-stram processing is presented. The best recognition results for both clean and noisy speech are obtained from the multi-stream combination of the J-Rasta-PLP features and the FF features.

Using morphossyntactic information in TTS systems: comparing strategies for European Portuguese

Abstract: To improve the quality of the speech produced by a Text-to-Speech (TTS) system, it is important to obtain the maximum amount of information from the input text that may help in this task. This covers a wide range of possibilities that can go from the simple conversion of non orthographic items to more complex syntactic and semantic analysis. In this paper, we present the development of a morphossyntactic tagging system and analyze its influence on the performance of a TTS system for European Portuguese.

Managing Authorship in Plot Conduction

Abstract: The increasing complexity of interactive applications and crescent demand for educational systems aggravates the urge for a balance between interactivity and scripted content. Approaches to this problem have ranged from a focus on interactivity and emotional engagement to linear narratives, with limited interactive capacities. The introduction of non-linearity frequently imposes a heavy burden to the authoring process. We argue that the definition of a domain ontology is the foundation that bonds all intervenients in the creative process. Furthermore, we propose a framework with which to cope the different issues involved in the development of interactive narrative applications. Going through a concrete example, we argue that these concepts and tools can contribute to a work methodology with which to attain good results within a feasible time frame.

Animated characters in bullying intervention

Abstract: The VICTEC (Virtual ICT with Empathic Characters) project explores the use of animated characters in virtual environments for educational issues, such as bullying behaviour. 76 participants evaluated a prototype of the VICTEC demonstrator. Results revealed high story believability with character conversation rated as convincing and interesting whilst character movement was poorly rated. The results imply that poor physical aspects of characters do not have detrimental effects on story believability and interest levels.

Adaptive nonlinear control of a distributed collector solar field

Abstract: This paper is concerned with an application of nonlinear adaptive techniques to temperature control in a distributed collector solar field. In the approach followed, the partial differential equation describing the field is approximated by a lumped parameter bi-linear model, whose states are the temperature values along the field. By using feedback exact linearisation together with a Lyapunov's approach, an adaptive controller is designed. This paper improves on previous work by using a better approximated model which takes into account that, in the field considered, temperature measures are only made at the input and at the output and not along the pipe. The design based on the improved simplified model allows faster convergence of parameter estimate and improved transient responde. The advantages of the new algorithm proposed are illustrated by means of simulations performed in a detailed physical model of the plant.

Fault Simulation Using Partially Reconfigurable Hardware

Abstract: This paper presents a fault simulation algorithm and that uses efficient partial reconfiguration of FPGAs. The methodology is particularly useful for evaluation of BIST effectiveness, and for applications in which multiple fault injection is mandatory, such as safety-critical applications. A novel fault collapsing methodology is proposed, which efficiently leads to the minimal stuck-at fault list at the look-up-tables’ terminals. Fault injection is performed using local partial reconfiguration with small binary files. Our results on the ISCAS’89 sequential circuit benchmarks show that our methodology can be orders of magnitude faster than software or fully reconfigurable hardware fault simulation..

The development of a multi-purpose spoken dialogue system.

Abstract: In this paper we describe a multi-purpose Spoken Dialogue System platform associated with two distinct applications as an home intelligent environment and remote access to information databases. These applications differ substantially on contents and possible uses but gives us the chance to develop a platform where we were able to represent diverse services to be accessible by a spoken interface. The implemented voice input/output possibilities and the service independence level opens a wide range of possibilities for the development of new applications using the current components of our Spoken Dialogue System.

RepWeb: replicated Web with referential integrity

Abstract: Replication of web content, through mirroring of web sites or browsing off-line content, is one of the most used techniques to increase content availability, reduce network bandwidth usage and minimize browsing delays in the world-wide-web.The world-wide-web does not support referential integrity, i.e., broken links do exist. This has been considered, for some years now, one of the most serious problems of the web. This is true in various fields, e.g.: i) if a user pays for some service in the form of web pages, he requires such pages to be reachable all the time, and ii) archived web resources, either scientific, legal or historic, that are still referenced, need to be preserved and remain available.Current approaches to the broken-link problem are not able to preserve referential integrity on the web and, simultaneously, support replication and minimize storage waste due to memory leaks. Some of them also impose specific authoring and management systems. Thus, the limitations of current systems reside in three issues: transparency, completeness and safety.We propose a system, RepWeb, comprised of an application to access and manage replicated web content and an implementation of an acyclic distributed garbage collection algorithm for wide-area replicated memory, that satisfies all these requirements. It supports replication, enforces referential integrity on the web and minimizes storage waste.

FantasyA - the duel of emotions

Abstract: FantasyA is a computer game where two characters face each other in a duel and emotions are used as the driving elements in the action decision of the characters. In playing the game, the user influences the emotional state of his or her semi-autonomous avatar using a tangible interface for affective input, the SenToy. In this paper we show how we approached the problem of modelling the emotional states of the synthetic characters, and how to combine them with the perception of the emotions of the opponents in the game. This is done by simulating the opponents action tendencies in order to predict their possible actions. For the user to play, he or she must understand the emotional state of his opponent which is achieved through animations (featuring affective body expressions) of the character. FantasyA was evaluated with 30 subjects from different ages and the preliminary results showed that the users liked the game and were able to influence the emotional states of their characters, in particular the young users.

A system for selective dissemination of multimedia information resulting from the alert project

Abstract: The media monitoring activity is undergoing a large expansion as a consequence of the different emerging media sources. This is pushing the development of automatic systems for selective dissemination of multimedia information. In this paper we present the development of a prototype system able to scan multimedia data, specifically TV broadcasts, and to generate alert messages to users about the relevant information to them. The system makes use of advanced processing technologies for content-based indexing of multimedia information. We use large vocabulary speech recognition system, associated with audio segmentation, and automatic topic indexing and segmentation, to generate category information as semantic markup of multimedia data. The system service is based on a web interface design offering new views of these marked documents and providing useful end-user services based on the content multimedia exploitation.

Topic indexing of TV broadcast news programs

Abstract: This paper describes a topic segmentation and indexation system for TV broadcast news programs spoken in European Portuguese. The system is integrated in an alert system for selective dissemination of multimedia information developed in the scope of an European Project. The goal of this work is to enhance the retrieval of specific spoken documents that have been automatically transcribed, using speech recognition. Our segmentation algorithm is based on simple heuristics related with anchor detection. The indexation is based on hierarchical concept trees (thesaurus), containing 22 main thematic domains, for which Hidden Markov models and topic language models were created. On-going experiments related to multiple topic indexing are also described, where a confidence measure based on the likelihood ratio test is used as the hypothesis test.

Machiavellian Characters and the Edutainment Paradox

Abstract: Purely script-based approaches to building interactive narratives have often limited interaction capabilities where variability demand exponential work. This is why Intelligent Virtual Agents (IVAs) are a transparent technique to handle user interaction in interactive narrative systems. However, it is hard to predict a sense of educational purpose in the global behavior of a group of IVAs, if no script or control is given. Efforts have been channelled to achieve such control, but are yet to achieve truly satisfactory results. These efforts are usually based on a direct connection between the control and the IVA architecture, which is a source of exponential complication. We propose a system that, based on a common ontology, can flexibly support the human authoring of educational goals independently of any specific IVA architecture. This is done by having a stage manager that follows an episodic based narrative where each episode is only specified through a set of properties and conditions that set the context for the characters. Although acting as they please, this contextualization will limit their range of action thereby facilitating the achievement of dramatic and educational goals.

Information sharing in mobile networks: a survey on replication strategies

Abstract: The emergence of more powerful and resourceful mobile devices, as well as new wireless communication technologies, is turning the concept of ad-hoc networking into a viable and promising possibility for ubiquitous information sharing. In such context, replication is a key issue in order to achieve acceptable performance and availability levels. However, the inherent characteristics of ad-hoc networks bring up new challenges for which most conventional replication systems don’t provide an appropriate response. Namely, the lack of a preexisting infrastructure, the high topological dynamism of these networks, the relatively low bandwidth of wireless links, as well as the limited storage and energy resources of mobile devices are issues that strongly affect the efficiency of any distributed system intended to provide ubiquitous information sharing. Such aspects demand solutions that are able to offer high availability, in spite of the expected frequent network partitions and device suspension periods. Pessimistic approaches for replication are usually too restrictive solutions to achieve such requirement. On the other hand, optimistic replication strategies offer weak consistency guarantees which may not reflect the expectations of users and applications. This paper aims at exposing the main challenges of a replicated system operating in the particular environment of mobile ad-hoc networks, with the goal of providing general guidelines for an effective solution. Based on a common conceptual model, some key design strategies are described and compared. A survey of some relevant state of the art solutions illustrates such concepts and presents implementations of the main design alternatives. In conclusion, guidelines are proposed towards an effective replication strategy for mobile ad-hoc networks.

A Data Mining Approach to Credit Risk Evaluation and Behaviour Scoring

Abstract: Behaviour scoring is used in several companies to score the customers according to credit risk by analyzing historical data about their past behaviour. In this paper we describe a data mining approach to credit risk evaluation in a Portuguese telecommunication company.

A Refereed Server for Educational CG Content

Abstract: Computer graphics has evolved considerably over the past few decades. As computer science, digital arts, and other areas of study that use computer graphics continue to evolve and gain new substance, educators have come to master new content and achieve deeper understandings of computers and imagery. As the core field becomes more mature, educators in all computer graphics disciplines have a greater need for high-quality curricular resources. Offering excellent educational materials is an important service to the community of educators. Such support will empower both young and seasoned educators alike to benefit from and contribute to the work of others. In this way, we can achieve a higher standard of teaching worldwide. The purpose of our work is to provide tools to foster such a community of computer graphics educators. We will present a system that will act as the means for their work to be appraised, assessed and made available to others through an online server for refereed educational content in computer graphics. In this paper we describe the basis for and highlight some of the starting requirements of CGEMS, the online Computer Graphics Educational Materials Server. This is organized around a web-based groupware application that supports the submission, review, acquisition, and archiving of curricular resources.

ASdeCopas: A Syntactic-Semantic Interface

Abstract: ASdeCopas is a syntactic-semantic parser, implemented in Prolog, which uses hierarchically organized order-independent rules. This paper focuses on the formalization of semantic rules, presenting the concepts of well-formed semantic rule, rules hierarchy, and the conditions for rules application. If two rules can apply, only the most specific one does so. Examples are given and some properties of the system are pointed out.

A 5.8-GHz Quadrature Cross-Coupled Oscillator

Abstract: This paper describes the implementation of a quadrature cross-coupled relaxation oscillator to be used in an OFDM RF front-end transceiver. A prototype of the oscillator was realized in a SiGe BiCMOS technology, and an oscillation frequency of 5.8 GHz was obtained which is 1/6 of the maximum fT of the bipolar transistors. The circuit performance is evaluated by simulation and by experiment.

Improving the accuracy of the speech synthesis based phonetic alignment using multiple acoustic features

Abstract: The phonetic alignment of the spoken utterances for speech research are commonly performed by HMM-based speech recognizers, in forced alignment mode, but the training of the phonetic segment models requires considerable amounts of annotated data. When no such material is available, a possible solution is to synthesize the same phonetic sequence and align the resulting speech signal with the spoken utterances. However, without a careful choice of acoustic features used in this procedure, it can perform poorly when applied to continuous speech utterances. In this paper we propose a new method to select the best features to use in the alignment procedure for each pair of phonetic segment classes. The results show that this selection considerably reduces the segment boundary location errors.

Algorithm for modulo (2/sup n/+1) multiplication

Abstract: An algorithm for designing efficient modulo (2n+1) multipliers based on Booth recoding is proposed With this algorithm, Wallace-tree adders can be used to design the fastest among all known modulo (2n+1) multipliers

A graph transformational approach to the multiprocessor scheduling of iterative computations

Abstract: A modular strategy for scheduling iterative computations is proposed. An iterative computation is represented using a cyclic task-graph. The cyclic task-graph is transformed into an acyclic task-graph. This acyclic task-graph is subsequently scheduled using one of the many well-known and high-quality static scheduling strategies from the literature. Graph unfolding is not employed and the generated schedules therefore require less memory than schedules generated through graph unfolding. Further, the number of iterations does not need to be known at compile-time. The effectiveness of the approach is compared to other methods including a graph unfolding strategy. We experimentally quantify how the task transformation affects the make-span of the schedules.

Gray Encoded Arithmetic Operators Applied to FFT and FIR Dedicated Datapaths.

Abstract: This paper addresses the use of low power techniques applied to FIR filter and FFT dedicated datapath architectures. New low power arithmetic operators are used as basic modules. In FIR filter and FFT algorithms, 2’s complement is the most common encoding for signed operands. We use a new architecture for signed multiplication, which maintains the pure form of an array multiplier. This architecture uses radix-2m encoding, which leads to a reduction of the number of partial lines. Each group of m bits uses the Gray code, thus potentially further reducing the switching activity both internally and at the inputs. The multiplier architecture is applied to the DSP architectures and compared with the state of the art. Due to the characteristics of the FIR filter and FFT algorithms, which involve multiplications of input data with appropriate coefficients, the best ordering of these operations in order to minimize the power consumption in the implemented architectures is also investigated. As will be shown, the use of the low power operators with an appropriate choice of coefficients can contribute for the reduction of power consumption of the FIR and FFT architectures. Additionally, a new algorithm for the partitioning and ordering of the coefficients is presented. This technique is experimented in a Semi-Parallel architecture which enables speedup transformation techniques.

A Decision Support System for IST Academic Information.

Abstract: This article describes the Decision Support System (DSS) for Academic Information being developed at Instituto Superior Tecnico, the Engineering School of the Technical University of Lisbon. In Portuguese, this project has been given the acronym SADIA (Sistema de Apoio a Decisao da Informacao Academica). This paper focuses on the early phases of the DSS development process, i.e., the business requirements definition and the dimensional modelling. First, we show how the business requirements of the School drive the definition of the DSS dimensional model. Second, we detail the logical dimensional model for a selected business process, the IST Student Admission process. Third, the corresponding physical design decisions are reported. The results obtained from the three phases were successfully validated by business users.