Archivio della ricerca - Fondazione Bruno Kessler
Not a member yet
21227 research outputs found
Sort by
Design Optimization of RF MEMS-Driven Triangular Resonators with Sierpinski Geometry for Dual-Band Applications
Get PDFThis paper proposes a detailed design study of resonating high-frequency notch filters driven by RF MEMS switches and their optimization for dual-band operation in the X-Band. Microstrip configurations will be considered for single and dual-band applications.
An SPDT (single-pole-double-thru) switch composed of double-clamped ohmic microswitches has been introduced to connect triangular resonators with Sierpinski geometry, symmetrically placed with respect to a microstrip line to obtain a dual notch response.
Close frequencies or spans as wide as 2 GHz can be obtained depending on the internal complexity and the edge side. The internal complexity has been modified to introduce the possibility of using the same edge size for the frequency tuning of an elementary cell, maintaining a fixed footprint, and allowing coupled structures to implement high-frequency filters of the same size and variable operational frequencies. Preliminary experimental results have been obtained as a confirmation of the predicted device functionality
When Harry Meets Superman: The Role of The Interlocutor in Persona-Based Dialogue Generation
No full textEndowing dialogue agents with persona information has proven to significantly improve the consistency and diversity of their generations. While much focus has been placed on aligning dialogues with provided personas, the adaptation to the interlocutor’s profile remains largely underexplored. In this work, we investigate three key aspects: (1) a model’s ability to align responses with both the provided persona and the interlocutor’s; (2) its robustness when dealing with familiar versus unfamiliar interlocutors and topics, and (3) the impact of additional fine-tuning on specific persona-based dialogues. We evaluate dialogues generated with diverse speaker pairings and topics, framing the evaluation as an author identification task and employing both LLM-as-a-judge and human evaluations. By systematically masking or disclosing information about interlocutor, we assess its impact on dialogue generation. Results show that access to the interlocutor’s persona improves the recognition of the target speaker, while masking it does the opposite. Although models generalise well across topics, they struggle with unfamiliar interlocutors. Finally, we found that in zero-shot settings, LLMs often copy biographical details, facilitating identification but trivialising the task
Different Speech Translation Models Encode and Translate Speaker Gender Differently
No full textRecent studies on interpreting the hidden states of speech models have shown their ability to capture speaker-specific features, including gender. Does this finding also hold for speech translation (ST) models? If so, what are the implications for the speaker’s gender assignment in translation? We address these questions from an interpretability perspective, using probing methods to assess gender encoding across diverse ST models. Results on three language directions (English → French/Italian/Spanish) indicate that while traditional encoder-decoder models capture gender information, newer architectures—integrating a speech encoder with a machine translation system via adapters—do not. We also demonstrate that low gender encoding capabilities result in systems’ tendency toward a masculine default, a translation bias that is more pronounced in newer architectures
Search for a new scalar decaying into new spin-1 bosons in four-lepton final states with the ATLAS detector
No full textTowards high-resolution X-ray spectral imaging
No full textThis work presents the development and preliminary testing of a novel class of Hybrid Pixel Detectors (HPDs), built by coupling a low-noise, event-driven analog readout ASIC with a fine-pitch solid-state pixel sensor. Our new HPD builds upon XPOL-III, a readily-available cutting-edge 180 nm CMOS VLSI ASIC integrating over 100,000 pixels with fully analog readout at 50
pitch on a hexagonal grid, with an active area of
. We developed two versions of the hybrid device: one with 750
thick and 100
pixel pitch, Schottky-type CdTe sensor, and one with 300
thick and 50
pixel pitch silicon sensor. The assembled devices exhibit excellent performance with full single-photon sensitivity, demonstrating strong potential for advanced X-ray spectral imaging applications. Spatial and energy resolution measurement results for both the silicon and CdTe devices will be presented. These results open up exciting perspectives for the implementation of high-performance HPDs in various fields requiring precise X-ray imaging and spectroscopy. We will discuss the detailed performance metrics of the two devices and explore the implications of this technology for future developments in X-ray detection systems
Combining Wearable Proximity Sensing and Digital Time Diaries for Longitudinal Network Data Collection
No full textObserving how social interactions in nature change over time and space is a major challenge. This study introduces the SocialScope project, which aims to develop an innovative approach to longitudinal data collection on social networks by integrating three technologies.
The first is a novel dual-radio proximity wearable sensor combining Bluetooth and ultra-wideband (UWB) radios, capturing the distance between individuals over time, therefore allowing the reconstruction of proximity networks from these spatial and temporal patterns [1].
The second is a smartphone application called iLog, which combines user self-reported information (e.g., time diary) and data passively collected from smartphone sensors [2]. Digital diaries provide repeated snapshots of participants' daily behaviours, minimizing reliance on retrospective recall and improving ecological validity. Smartphone sensors, e.g., GPS, allow us to track subjects' daily activities and social and spatial interactions. The third is a close-ended questionnaire.
Here we explore some early evidence on how people adjust interpersonal distances in everyday contexts based on gender, nationality, and age attributes. Eighteen college students living on the same dormitory floor wore the proximity sensors for 15 days, and continuously measured distances during real-life encounters every 15 seconds. At the same time, the app installed on the subjects' smartphones gathered time diary information every 30 minutes (Where are you? Who are you with? What are you doing? What is your mood?) and continuous information from the smartphone's sensors. The analysis is based on a total of 431,329 distances detected with the tag, with a total of 6,299 interactions lasting 90 seconds or longer, and 4,681 self-reported time diary information
Proceedings of the 3rd Workshop on Gender-Inclusive Translation Technologies (GITT 2025)
No full textL’enigma di una fede resiliente. Davide Zordan e il futuro della teologia
No full textIl saggio è un tributo a Davide Zordan e alla sua opera a dieci anni dalla prematura scomparsa. Dopo una rapida rassegna del suo itinerario intellettuale, l’analisi si concentra sulla sua visione della teologia come intelligenza critica della fede, in cui immaginazione e sensibilità svolgono un ruolo fondamentale. L’enfasi sulla concretezza dell’esperienza di fede spiega anche il secondo elemento che viene preso in esame nell’articolo: la centralità epistemica della collaborazione tra teologia e cultural studies. Il saggio si conclude con il tentativo di ricavare dagli ultimi scritti di Zordan un’immagine ambiziosa e fiduciosa della teologia del futuro
Echoes of Phonetics: Unveiling Relevant Acoustic Cues for ASR via Feature Attribution
No full textDespite significant advances in ASR, the specific acoustic cues models rely on remain unclear. Prior studies have examined such cues on a limited set of phonemes and outdated models. In this work, we apply a feature attribution technique to identify the relevant acoustic cues for a modern Conformer-based ASR system. By analyzing plosives, fricatives, and vowels, we assess how feature attributions align with their acoustic properties in the time and frequency domains, also essential for human speech perception. Our findings show that the ASR model relies on vowels’ full time spans, particularly their first two formants, with greater saliency in male speech. It also better captures the spectral characteristics of sibilant fricatives than non-sibilants and prioritizes the release phase in plosives, especially burst characteristics. These insights enhance the interpretability of ASR models and highlight areas for future research to uncover potential gaps in model robustness
Dual-Frequency Radar Sounder Correlation Measurements For Surface Roughness Mapping
No full textPlanetary radar sounders provide information that helps to characterize the surface and subsurface of planetary bodies, providing insights into their structure, composition, and topography. Surface roughness is one of the main interesting variables that can be estimated by radar sounder data. In this work, we propose a technique for generating surface roughness maps with an imaging technique applied to dual-frequency radar sounder data. The method projects the radar echoes onto a digital elevation model (DEM) and combines simultaneous acquisitions at two frequency bands. Exploiting the omnidirectional nature of dipole antennas, we integrate signals from multiple range lines and orbit tracks to resolve ambiguities in the direction of arrival. The result is a cross-correlation metric derived from projections of dual-frequency acquisitions, representing the probed surface roughness. We demonstrated this approach using MARSIS multifrequency radar sounder data acquired over Mars. In particular, we generated a roughness map of the Olympia Undae region. The method confirmed its effectiveness as it is correlated with other Martian surface roughness maps and presents a sensibility of 50∼100 m-scale roughness features