You are here


Primary tabs

  1. Edmondo Trentin; Marco Gori,
    A Survey of Hibrid ANN/HMM Models for Automatic Speech Recognition,
    In spite of the advances accomplished throughout the last decades by a number of research teams, Automatic Speech Recognition (ASR) is still a challenging and difficult task. In particular, recognition systems based on hidden Markov models (HMMs) are effective under many circumstances, but do suffer from some major limitations that limit applicability of ASR technology in real-world environments. Attempts were made to overcome these limitations with the adoption of Artificial Neural Networks (ANN) as an alternative paradigm for ASR, but ANN were unsuccessful in dealing with long time-sequences of speech signals. Between the end of the Eighties and the beginning or the Nineties, some researchers began exploring a new research area, by combining HMMs and ANNs within a single, hybrid architecture. The goal in hybrid systems for ASR is to take advantage from the properties of both HMMs and ANNs, improving flexibility, hardware implementability and recognition performance. A variety of different architectures and novel training algorithms have been proposed in literature. This paper reviews a number of significant hybrid models for ASR. Early attempts to emulate HMMs by ANNs are briefly described. Them we focus on ANNs to estimate posterior probabilities of the states of an HMM and on 'global' optimization, where a single, overall training criterion is defined over the HMM and the ANNs, Connectionist Vector Quantization for discrete HMMs, and other more recent approaches are also reviewed, It is pointed out that, in addition to their theoretical interest, hybrid systems have been allowing for tangible improvements in recognition performance over the standard HMMs in difficult and significant benchmark tasks,
  2. Marco Aste; Massimo Boninsegna; M. Della Torre,
    Finding Perspective Projection and Stereo Localization mappings for a Multi-Camera System,
    This paper investigates the problem of finding the perspective projection and the stereo localization transforms for a binocular imaging system with long baseline. Neural techniques are used to estimate the geometrical mappings from a set of associations between points spread in the 3D space and their corresponding image projections. Aim of this work is to explore how neural networks can deal with acquisition noise and optical distorsions without considering complex camera models. Three techniques are experimentally compared: one based on the simple pin-hole camera model, a second purely based on neural function approximation, and a third which uses neural networks only to account for the deviations of the actual data from the pin-hole model predictions. Experiments have been performed on real data collected by a four-camera system placed in a laboratory room. Performance has been measured in terms of Euclidean distance between actual targets and estimated outputs on a validation and a test set. The results show the effectiveness of the neural approach and validate the combined use of pin-hole model and neural nets to a large extent. The generalization ability of the neural architectures has been finally investigated on a empirical basis,
  3. Roldano Cattoni; Alessandra Potrich,
    Bayesian Belief Networks: Introduction and Learning,
    Bayesian Belief Networks are graph-based representations of probability distributions. In the last decade they became popular for modeling and using uncertain knowledge in many and different contexts. In this paper an introduction to the framework and a review of the main issues related to learning Bayesian Belif Networks are presented. The first part focuses on the definition of the framework: the mathematical and representational properties are described and discussed as well as Belief Networks from data, a topic which received much attention recently. A large amount of works, approaches and methodologies proposed in the literature is surveyed,
  4. Francesco Ricci; Paolo Avesani,
    Workshop Italiano sul Ragionamento basato su Casi,
  5. Lorenzo Potrich,
    Metodi visivi per la caratterizzazione automatica del moto di bolle di aria in acqua,
  6. Alessandro Cimatti; Roberto Sebastiani,
    Servizi forniti dagli strati Safety Layer e Connection Manager,
    Obiettivo del presente documento è fornire una (proposta preliminare di) descrizione dei servizi forniti agli applicativi dagli strati Connection Manager (CM) e Safety Layer (SL), come perzialmente descritto nei documenti (1) e (3). Tale proposta è risultante della riunione in Ansaldo del 08-05-98 tra Ansaldo e Irst,
  7. Alessandro Cimatti; Roberto Sebastiani,
    Specifica formale dei protocolli Safety Layer e Connection Manager,
    Il presente documento descrive una specifica formale del protocollo di comunicazione tra applicativi in sicurezza basato su ridondanza. Il protocollo è stato suddiviso in due livelli separati, il Safety Layer (SL), come parzialmente descritto nei documenti (1) e (3), ed il Connection Manager (CM). Il S: e il CM sono stati specificati formalmente in SDL, e sono state applicate tecniche di validazione basate su simulazione esaustiva (model checking) attraverso il tool ObjectGEODE Nel presente documento vengono descritti i servizi forniti da tali livelli, e viene descritta la specifica formale delle macchine a stati che realizzano i livelli CM e SL. i dettagli del codice SDL sono descritti nel documento (6).,
  8. Alessandro Cimatti; Roberto Sebastiani,
    Protocolli Safety Layer e Connection Manager: Descrizione del Codice SDL,
    Nel presente documento viene descritto in dettaglio il codice SDL della specifica formale dei protocolli Safety Layer (SL) e Connection Manager (CM) descritta in (1). Il modello è realizzato attraverso i meccanismi di 'definizione di tipo' e di 'istanziazione' di SDL. Il codice è ripartito nei fine, contenente la descrizione dei tipi (es. macchina SL, macchina CM), e, contentenente le istanze che realizzano la particolare configurazione analizzata Il presente documento assume che il lettore abbia una certa conoscenza di SDL, e fa esplicito riferimento al listato dei file e allegato.,
  9. Emanuele Pianta; L. Tovena,
    Generating with flexible templates from C-STAR Interchange Format,
    We present a system for generating Italian sentences from the interlingua semantic representation (Interchange Format) adopted within the C-STAR speech to speech translation project. The generation task in our application scenario is made peculiar by: i) a semantically underspecified input representation whose interpretation relies on implicit domain knowledge; ii) the spoken language output, characterized by frequent use of idiomatic forms, fragmentary phrases, etc. iii) a strong requirement for high time efficiency. We discuss how these constraints lead us to develop a template based generator providing a good trade-off betzeen flexibility and efficiency. We overcome the shortcomings of traditional static templates by using flexible templates which allow an elegant treatment of phonological adjustments, morphological agreement (very frequent in Italian) and syntactic constituency phenomena,
  10. Carla Maria Modena; Stefano Messelodi,
    Progetto CODICE*: Alcune considerazioni sulle modalità di valutazione del sistema,
    Queste note raccolgono alcune considerazioni relative alla definizione delle modalità di valutazione del sistema di analisi ed interpretazione di documenti complessi in via di sviluppo nell’ambito del progetto CODICE. Viene messo in evidenza che la possibilità di definire metodi di valutazione generali è attuabile solo per alcune sottoparti del sistema se non per specifiche procedure, mentre la definizione di una validazione globale richiede che sia fissato lo scenario applicativo,