Ricerc@Sapienza

Graph-based representations play a key role in machine learning. The fundamental step in these representations is the association of a graph structure to a dataset. In this paper, we propose a method that finds a block sparse representation of the data by associating a graph, whose Laplacian matrix admits the sparsifying dictionary as its eigenvectors. The main idea is to associate a graph topology to the data in order to make the observed signals band-limited over the inferred graph.

This paper introduces innovative detection schemes for Forward Scatter Radar (FSR) based on the GLRT (Generalized Likelihood Ratio Test) for both cases, where a fixed threshold can be used and where a fully adaptive CFAR scheme is desired. The detection performance of the newly proposed detectors is characterized analytically and compared to the performance of the standard detection scheme. This shows that the new detectors always outperform the standard FSR detector.

The forward scatter radar (FSR) configuration is especially appealing for the detection of low-observable targets since it provides well-known properties for the radar cross section, which includes enhancements with respect to the monostatic and/or moderate bistatic configurations, aswell as robustness to the target material and detailed geometrical characteristics.

The rapid development of smart devices, such as smartphones and tablets, leads to new challenges and ushers in a new stage of human-computer interaction. In this context, it becomes essential to develop methods and techniques for a better and more natural interaction with these devices. In this article, we address the problem of gesture segmentation and recognition, taking into account the limited computational resources of smartphone devices.

In this paper, a new model of the frequency dependence of the double-barrier THz rectifier is presented. The new structure is of interest because it can be realized by CMOS image sensor technology. Its application in a complex field such as that of THz receivers requires the availability of an analytical model, which is reliable and able to highlight the dependence on the parameters of the physical structure. The model is based on the hydrodynamic semiconductor equations, solved in the small signal approximation.

To improve the gate oxide long term performance of MOSFETs in back side illuminated CMOS image sensors the wafer back is patterned with suitable through-silicon-trenches. We demonstrate that the reliability improvement is due to the annealing of the gate oxide border traps thanks to passivating chemical species carried by trenches.

Motivated by recent theoretical challenges for 5G, this study aims to position relevant results in the literature on codedomain non-orthogonal multiple access (NOMA) from an information-theoretic perspective, given that most of the recent intuition of NOMA relies on another domain, that is, the power domain. Theoretical derivations for several code-domain NOMA schemes are reported and interpreted, adopting a unified framework that focuses on the analysis of the NOMA spreading matrix, in terms of load, sparsity, and regularity features.

Aim of the present study was to test and validate an ex vivo predictive model for the evaluation of the shrinkage occurring in hepatic tissue during a microwave thermal ablation procedure. Microwave ablation (N=134) was conducted with three different commercial devices on cubes of ex vivo liver (15-40\pm 2 mm side) embedded in agar phantoms. 50-60W was applied for 1-10 min duration. Pre-and post-ablation dimensions of the samples, as well as the extent of carbonization and coagulation were measured.

The paper proposes two different Tissue Mimicking Material (TMM) techniques for the development of breast phantoms which are suitable for multi-modality imaging. In particular, the focus is on the behavior of dielectric and acoustic properties when fat, sodium chloride and sugar are added to the mixtures.

This communication describes the ongoing efforts towards the assessment of microwave imaging as a tool for real-time monitoring of thermal ablation treatments. In particular, the ex-vivo experimental set-up adopted for the validation is described, and the results of a preliminary experiment are shown. Notably, by analyzing pre- and post-ablation treatment data it is possible to recognize the footprint of the interface between the ablated and not-ablated tissue, making it possible to estimate the boundary of the treated area.