The goal of this project is to explore new theoretical tools for processing observations over a networked domain and to apply them to cutting edge problems in information and communication. Signal processing on networks is deemed relevant since i) it improves processing performance by natural integration of prior information, ii) it extends applicability of artificial intelligence tools to non Euclidean data domains, iii) it matches novel kind of data structures which are brought in evidence by emerging fields of study. The project will explore novel filtering approaches for signals defined on graphs/manifolds, the application of these approaches to compressed representation and transmission of extended reality video content, involving natural and synthetic content. Finally, the project research framework will be applied to the relevant topic of biological network analysis, with particular reference to processing real biological network data, with particular reference, but not limited to, brain connectivity networks .
The project relies on strong premises, for the scientific interest of the general background, for the novelty of the proposed themes, as well as for the complementary expertise and consolidated collaboration of the participants on topic related to the project, which has already led to preliminary promising results. As for the general approach, there is an increasing interest to exploit the potentiality of signal processing on networked domain. This occurs mainly because of the availability of new data acquisition system where the networked domain is inherited by the nature of the observations. The boosting research on the project topic is reflected by the appearance of several new works, and even of novel journals on this approach. The integrated signal/network processing is novel, because, although graph study dates back to several decades ago, the point of view of signal over network processing is relatively new (e.g. the IEEE Transactions on Signal Processing and IEEE/ACM Transactions On Networking date back to 1953 and 1993, respectively, while the IEEE Transactions on Signal and Information Processing over Networks is just five years old).
As for the novelty of the proposed topics, namely i) the definition of manifold processing operators measuring the local variations of the signal defined on graph, ii) the application of the operators to the design of extended reality streaming services, and iii ) their application to inference and classification problems solving on biomedical data (EEG), we report in Figs. 4, 5, 6 above the main related works. From these bibliographic summaries we recognize the huge interest of the project topic in cutting edge literature. On the other hand, the suggested guidelines for the project development have strong theoretical ground and wide applicability, and all premises for the production of novel relevant contributions to the field.
As for the project team members, they have already cooperated on research that establishes the ground for development of new studies. As for the filtering, past research contributions of the participants on one hand have dealt with the adoption of CHF filtering as a way to extract visually relevant information in image recovery problems [Pa03], while on the other hand have investigated algebraic description of the connectivity properties [Ab11]. Both the paths of investigation have achieved an excellent impact. As for the design of advanced streaming services, the participants have already published relevant papers on the topics, mostly with international co-authors. As for the real data application fields, the participants to the project have matured experiences, either through a PhD project or by collaboration started within the STITCH Sapienza center. Preliminary results have been submitted to relevant journals [Cat19], [Col20]. Thereby, the vision achieved as an outcome of previous research together with the novelty of the actual project content pave the way to novel relevant contributions on the project topic.
[Ab11] Abbagnale, A. and Cuomo, F., 2011. Leveraging the algebraic connectivity of a cognitive network for routing design. IEEE transactions on mobile computing, 11(7), pp.1163-1178.
[Pa03] Panci, G., Campisi, P., Colonnese, S. and Scarano, G., 2003. Multichannel blind image deconvolution using the bussgang algorithm: Spatial and multiresolution approaches. IEEE Transactions on Image Processing, 12(11), pp.1324-1337.
[Col20] Colonnese, S. , Di Lorenzo, P. , Cattai, T., Scarano, G., De Vico Fallani, F., ¿A Joint Markov Model for Communities, Connectivity and Signals defined over Graphs¿, accepted with mandatory minor revisions at IEEE Signal Processing Letters
[Cat19] Cattai, Tiziana, Stefania Colonnese, Marie-Constance Corsi, Danielle S. Bassett, Gaetano Scarano, and Fabrizio De Vico Fallani. "Phase/amplitude synchronization of brain signals during motor imagery BCI tasks." arXiv preprint arXiv:1912.02745 (2019).