LEARNING OUTCOMES

- To introduce theory and techniques for architectural design of context-aware telecommunications systems able to provide informative services according to a cognitive paradigm

- To provide a common framework to identify and to describe methodologies and techniques for perception, representation and analysis of contextual multisensorial physical (radio, video, audio, etc.) and virtual signals (e.g.network-based context data))

- To provide a common framework to identify and to describe methodologies and techniques for integrating multisensorial contextual data by using Data Fusion paradigms and techniques

- To provide a common framework for defining behavioral artificial models for context based, adaptive and personalized  decision steps used by cognitive system to address and react with respect to different contextual working situations.

- To show examples and applications of specific techniques within cognitive telecommunication systems by means of description of two main case studies: cognitive radio and multisensor/multimodal cognitive human-machine interfaces in smart spaces. 

TEACHING METHODS

Lectures and lab exercises Project based learning

SYLLABUS/CONTENT

1) INTRODUCTION  General definitions and models for cognitive systems. Behavioral cognitive artificial models for context based, adaptive and personalized  decision The cognitive cycle model; perception, analysis, decision, action. Logical and bio-inspired cognitive system models. Cognitive Data fusion functional architectural the JDL model and its extensions. Haykin-Fuster Cognitive Dynamic Systems. The Probabilistic Graphical Model based Data fusion architecture

 2) Data Fusion methodologies and techniques for integrating multisensorial contextual data

Acquisition  and representation of contextual data.. Contextual data hierarchical representation: presence, localization, behavior, situation, threat.  Methodologies and techniques for physical sensor signal processing: digital signal processing issues with radio, video, audio signals. Techniques and algorithms for acquisition and analysis of contextual data. Bayesian Data Fusion processing techniques: alignment, data association, state estimation, abnormality detection   Probabilistic Graphical Models: Dynamic Bayesian Networks (DBN)  and Markov Random Fields. Representation and inference. Factorization and Belief propagation. Bayesian State estimation techniques: Kalman filter, Extended Kalman Filter, Unscented Kalman Filter, Particle Filtering.  Processing algorithms and PGM representation   Situation assessment: Interaction Model Representation using coupled DBNs. State, superstate and event based interaction representation. Methods for dimensionality reduction and classification: elf Organizing Map, Neural Gas. Threat assessment by incremental evaluation of  distance between Prediction from Update in a Bayesian node.  Kllback Leiber. Distributed decision theory.

 3) Case studies

1) design of  cognitive Data fusion systems with application to health, surveillance, smart environments, robotsic Lego applications.

RECOMMENDED READING/BIBLIOGRAPHY

 Basic: Class notes written  by the lecturer and made available through Internet