arXiv subjects

Michele Fiori

Publications and source records attributed to Michele Fiori.

18 recordsLinked to original sources

arXiv2026

Kilometre-scale Jovian moon characterized for a potential JUICE flyby

Jupiter's irregular satellites are thought to be relics of early Solar System planetesimals. However, their small sizes, large distances from Earth and close angular proximity to Jupiter make them difficult to characterize remotely. Here we report multi-instrument observations of kilometre-sized Kallichore, the only irregular moon of Jupiter amenable to a close flyby by ESA's Jupiter Icy Moons Explorer (JUICE), whose astrometric and physical properties were still poorly constrained. We used Hubble photometry and astrometry, followed by a ground-based, multi-site stellar occultation campaign and astrometric and photometric observations with the 10.4-m Gran Telescopio Canarias. We reduced Kallichore's orbital uncertainty by up to approximately 80% and determined its shape: an elongated object with a minimum semi-axis ratio of a/b = 1.53 +/- 0.10, an area-equivalent diameter of 3.8 (+2.3/-0.3) km and a dark surface with a geometric albedo of 3.7% (+0.7/-2.2). No close companions were detected. These new constraints provide a viable pathway towards a JUICE flyby once the spacecraft arrives in the Jupiter system in 2031.

astro-ph.EP↗

arXiv2026

IQUEYE at Gemini South: instrument, science commission, and first results

The Italian quantum eye (IQUEYE) is a fast photon counter based on the single photon avalanche diode detectors and capable of preserving a ~0.5 ns/h accuracy photon time of arrival. IQUEYE was originally developed for intensity interferometry experiments, but now its scientific scope has been extended towards ultra fast astronomy, including optical pulsars, millisecond pulsars and the enigmatic fast radio bursts. IQUEYE's capabilities are mainly restricted by the number of photons detected, a quantity that scales with the collector size of an optical telescope. Through the visitor instrument program at Gemini South (Cerro Pachón, Chile) we brought IQUEYE to the 8.1-m dish, reaching an order magnitude sensitivity increased from previous operations. At Gemini South we installed IQUEYE to observe giant pulse emitters, millisecond pulsars, and transitional millisecond pulsars for over 40 hours in the span of a week.

astro-ph.IM↗

arXiv2026

Evolution of the transitional millisecond pulsar PSR J1023+0038 from Aqueye+ and NICER observations

Transitional millisecond pulsars (tMSPs) are old neutron stars spun up by accretion from a low-mass companion. These objects can switch between two emission regimes: rotation-powered radio pulsar and accreting X-ray pulsar. The origin of their optical and X-ray pulsations is still debated, although one model attributes them to synchrotron emission produced in a shock between the pulsar wind and the accretion flow. The small phase lag observed between optical and X-ray pulses in PSR J1023+0038 supports a common origin. We present a new measurement of the phase lag between optical and X-ray pulse profiles of PSR J1023+0038 and investigate the evolution of the time of passage at the ascending node ($T_{\rm{asc}}$) up to 2023. We performed a timing analysis of optical observations obtained with Aqueye+ between 2021 and 2023 and of X-ray data from NICER in 2023. We derive updated values of $T_{\rm{asc}}$ and measure the optical - X-ray phase lag from simultaneous observations. We find that $T_{\rm{asc}}$ increases by about 20 s per year. In January 2023, we measure a phase lag of $0.067 \pm 0.018$, corresponding to $112.3 \pm 30.7\,μ$s. Since 2017, the evolution of $T_{\rm{asc}}$ follows a parabolic trend, indicating an increase in the orbital period and orbital separation of the system. This behaviour is consistent with non-conservative Roche-lobe overflow, with the donor losing mass at a rate much higher than the accretion rate. The phase lag measurement further supports a common origin of the optical and X-ray pulsations.

astro-ph.HE↗

arXiv2026

DomusFM: A Foundation Model for Event-Based Behavioral Monitoring in Smart-Homes

Smart-home sensor-based behavioral monitoring holds significant potential for healthcare, independent living, and early detection of functional or cognitive changes. In this setting, tasks like activity recognition, prediction, and pattern discovery provide complementary views of daily life, supporting the modeling of personal routines and habits, and their long-term changes. Existing approaches, however, face critical limitations. Supervised approaches require impractical amounts of labeled activity data to capture the variability of daily behavior across residents and environments. Foundation models represent a promising direction for learning transferable representations of latent behavioral patterns from sensor data. Still, current efforts are mostly designed for continuous inertial or physiological sensor data and do not address the sparse, discrete, and semantically rich event streams produced by smart homes. In this paper, we introduce DomusFM, a domain-specific foundation model for sensor-based behavioral monitoring in smart homes based on semantic event streams. DomusFM employs a self-supervised dual contrastive learning paradigm to capture both event-level semantic attributes and sequence-level temporal dependencies. By integrating semantic embeddings from a lightweight language model and specialized encoders for temporal patterns and binary states, DomusFM learns transferable representations that can be adapted across heterogeneous smart-home environments and tasks related to activity and event analysis. Through a leave-one-dataset-out evaluation across seven public smart-home datasets, we demonstrate that DomusFM consistently outperforms baselines on three downstream tasks: ADL recognition, next-k event prediction, and unsupervised clustering. DomusFM has a small footprint and can be deployed on edge devices.

cs.AI↗

arXiv2026

Improving Zero-shot ADL Recognition with Large Language Models through Event-based Context and Confidence

Unobtrusive sensor-based recognition of Activities of Daily Living (ADLs) in smart homes by processing data collected from IoT sensing devices supports applications such as healthcare, safety, and energy management. Recent zero-shot methods based on Large Language Models (LLMs) have the advantage of removing the reliance on labeled ADL sensor data. However, existing approaches rely on time-based segmentation, which is poorly aligned with the contextual reasoning capabilities of LLMs. Moreover, existing approaches lack methods for estimating prediction confidence. This paper proposes to improve zero-shot ADL recognition with event-based segmentation and a novel method for estimating prediction confidence. Our experimental evaluation shows that event-based segmentation consistently outperforms time-based LLM approaches on complex, realistic datasets and surpasses supervised data-driven methods, even with relatively small LLMs (e.g., Gemma 3 27B). The proposed confidence measure effectively distinguishes correct from incorrect predictions.

cs.CV↗

arXiv2026

GNN-XAR: A Graph Neural Network for Explainable Activity Recognition in Smart Homes

Sensor-based Human Activity Recognition (HAR) in smart home environments is crucial for several applications, especially in the healthcare domain. The majority of the existing approaches leverage deep learning models. While these approaches are effective, the rationale behind their outputs is opaque. Recently, eXplainable Artificial Intelligence (XAI) approaches emerged to provide intuitive explanations to the output of HAR models. To the best of our knowledge, these approaches leverage classic deep models like CNNs or RNNs. Recently, Graph Neural Networks (GNNs) proved to be effective for sensor-based HAR. However, existing approaches are not designed with explainability in mind. In this work, we propose the first explainable Graph Neural Network explicitly designed for smart home HAR. Our results on two public datasets show that this approach provides better explanations than state-of-the-art methods while also slightly improving the recognition rate.

cs.AI↗

arXiv2025

Leveraging Large Language Models for Explainable Activity Recognition in Smart Homes: A Critical Evaluation

Explainable Artificial Intelligence (XAI) aims to uncover the inner reasoning of machine learning models. In IoT systems, XAI improves the transparency of models processing sensor data from multiple heterogeneous devices, ensuring end-users understand and trust their outputs. Among the many applications, XAI has also been applied to sensor-based Activities of Daily Living (ADLs) recognition in smart homes. Existing approaches highlight which sensor events are most important for each predicted activity, using simple rules to convert these events into natural language explanations for non-expert users. However, these methods produce rigid explanations lacking natural language flexibility and are not scalable. With the recent rise of Large Language Models (LLMs), it is worth exploring whether they can enhance explanation generation, considering their proven knowledge of human activities. This paper investigates potential approaches to combine XAI and LLMs for sensor-based ADL recognition. We evaluate if LLMs can be used: a) as explainable zero-shot ADL recognition models, avoiding costly labeled data collection, and b) to automate the generation of explanations for existing data-driven XAI approaches when training data is available and the goal is higher recognition rates. Our critical evaluation provides insights into the benefits and challenges of using LLMs for explainable ADL recognition.

cs.CL↗

arXiv2025

The SERENADE project: Sensor-Based Explainable Detection of Cognitive Decline

Mild Cognitive Impairment (MCI) affects 12-18% of individuals over 60. MCI patients exhibit cognitive dysfunctions without significant daily functional loss. While MCI may progress to dementia, predicting this transition remains a clinical challenge due to limited and unreliable indicators. Behavioral changes, like in the execution of Activities of Daily Living (ADLs), can signal such progression. Sensorized smart homes and wearable devices offer an innovative solution for continuous, non-intrusive monitoring ADLs for MCI patients. However, current machine learning models for detecting behavioral changes lack transparency, hindering clinicians' trust. This paper introduces the SERENADE project, a European Union-funded initiative that aims to detect and explain behavioral changes associated with cognitive decline using explainable AI methods. SERENADE aims at collecting one year of data from 30 MCI patients living alone, leveraging AI to support clinical decision-making and offering a new approach to early dementia detection.

cs.LG↗

arXiv2025

Large Language Models are Zero-Shot Recognizers for Activities of Daily Living

The sensor-based recognition of Activities of Daily Living (ADLs) in smart home environments enables several applications in the areas of energy management, safety, well-being, and healthcare. ADLs recognition is typically based on deep learning methods requiring large datasets to be trained. Recently, several studies proved that Large Language Models (LLMs) effectively capture common-sense knowledge about human activities. However, the effectiveness of LLMs for ADLs recognition in smart home environments still deserves to be investigated. In this work, we propose ADL-LLM, a novel LLM-based ADLs recognition system. ADLLLM transforms raw sensor data into textual representations, that are processed by an LLM to perform zero-shot ADLs recognition. Moreover, in the scenario where a small labeled dataset is available, ADL-LLM can also be empowered with few-shot prompting. We evaluated ADL-LLM on two public datasets, showing its effectiveness in this domain.

cs.AI↗

arXiv2025

ContextGPT: Infusing LLMs Knowledge into Neuro-Symbolic Activity Recognition Models

Context-aware Human Activity Recognition (HAR) is a hot research area in mobile computing, and the most effective solutions in the literature are based on supervised deep learning models. However, the actual deployment of these systems is limited by the scarcity of labeled data that is required for training. Neuro-Symbolic AI (NeSy) provides an interesting research direction to mitigate this issue, by infusing common-sense knowledge about human activities and the contexts in which they can be performed into HAR deep learning classifiers. Existing NeSy methods for context-aware HAR rely on knowledge encoded in logic-based models (e.g., ontologies) whose design, implementation, and maintenance to capture new activities and contexts require significant human engineering efforts, technical knowledge, and domain expertise. Recent works show that pre-trained Large Language Models (LLMs) effectively encode common-sense knowledge about human activities. In this work, we propose ContextGPT: a novel prompt engineering approach to retrieve from LLMs common-sense knowledge about the relationship between human activities and the context in which they are performed. Unlike ontologies, ContextGPT requires limited human effort and expertise. An extensive evaluation carried out on two public datasets shows how a NeSy model obtained by infusing common-sense knowledge from ContextGPT is effective in data scarcity scenarios, leading to similar (and sometimes better) recognition rates than logic-based approaches with a fraction of the effort.

cs.LG↗

arXiv2024

Using Large Language Models to Compare Explainable Models for Smart Home Human Activity Recognition

Recognizing daily activities with unobtrusive sensors in smart environments enables various healthcare applications. Monitoring how subjects perform activities at home and their changes over time can reveal early symptoms of health issues, such as cognitive decline. Most approaches in this field use deep learning models, which are often seen as black boxes mapping sensor data to activities. However, non-expert users like clinicians need to trust and understand these models' outputs. Thus, eXplainable AI (XAI) methods for Human Activity Recognition have emerged to provide intuitive natural language explanations from these models. Different XAI methods generate different explanations, and their effectiveness is typically evaluated through user surveys, that are often challenging in terms of costs and fairness. This paper proposes an automatic evaluation method using Large Language Models (LLMs) to identify, in a pool of candidates, the best XAI approach for non-expert users. Our preliminary results suggest that LLM evaluation aligns with user surveys.

cs.HC↗

arXiv2024

Towards Image Synthesis with Photon Counting Stellar Intensity Interferometry

Stellar intensity interferometry (SII) is based on the correlation of the light intensity fluctuations of a star detected at two or more telescopes, with no need to combine the collected photons directly. A measurement of the correlation in full "photon-counting mode" was experimented with fast photon counters in Italy (2016-2020) and is currently being adapted to the ASTRI Mini-Array. Performing image synthesis with "photon-counting" SII requires a series of preparatory activities that involve the optimization of the pipelines for the treatment of time series acquired at extremely high photon rates, the development of efficient and innovative algorithms for the cross-correlation of the arrival times in large time series and the development of a preliminary version of a dedicated pipeline for the synthesis of images starting from interferometric data. Here we present the project and the present status of the activities.

astro-ph.IM↗

arXiv2022

Investigating the accuracy achievable in reconstructing the angular sizes of stars through stellar intensity interferometry observations

Context: In recent years, stellar intensity interferometry has seen renewed interest from the astronomical community because it can be efficiently applied to Cherenkov telescope arrays. Aims: We have investigated the accuracy that can be achieved in reconstructing stellar sizes by fitting the visibility curve measured on the ground. The large number of expected available astronomical targets, the limited number of nights in a year, and the likely presence of multiple baselines will require careful planning of the observational strategy to maximise the scientific output. Methods: We studied the trend of the error on the estimated angular size, considering the uniform disk model, by varying several parameters related to the observations, such as the total number of measurements, the integration time, the signal-to-noise ratio, and different positions along the baseline. Results: We found that measuring the value of the zero-baseline correlation is essential to obtain the best possible results. Systems that can measure this value directly or for which it is known in advance will have better sensitivity. We also found that to minimise the integration time, it is sufficient to obtain a second measurement at a baseline half-way between 0 and that corresponding to the first zero of the visibility function. This function does not have to be measured at multiple positions. Finally, we obtained some analytical expressions that can be used under specific conditions to determine the accuracy that can be achieved in reconstructing the angular size of a star in advance. This is useful to optimise the observation schedule.

astro-ph.IM↗

arXiv2022

Deep upper limit on the optical emission during a hard X-ray burst from the magnetar SGR J1935+2154

In September 2021 the magnetar SGR J1935+2154 entered a stage of burst/flaring activity in the hard X-ray band. On September 10, 2021 we observed SGR J1935+2154 with the fiber-fed fast optical photon counter IFI+Iqueye, mounted at the 1.22 m Galileo telescope in Asiago. During one of the IFI+Iqueye observing windows a hard X-ray burst was detected with the Fermi Gamma-ray Burst Monitor. We performed a search for any significant increase in the count rate on the 1-s, 10-ms and 1-ms binned IFI+Iqueye light curves around the time of the Fermi burst. No significant peak was detected with a significance above 3$σ$ in an interval of $\pm$90 s around the burst. Correcting for interstellar extinction ($A_V \simeq 5.8$ mag), the IFI+Iqueye upper limits to any possible optical burst from SGR J1935+2154 are $V=10.1$ mag, $V=7.2$ mag and $V=5.8$ mag for the 1-s, 10-ms and 1-ms binned light curves, respectively. The corresponding extinction corrected upper limits to the fluence (specific fluence) are $3.1 \times 10^{-10}$ erg cm$^{-2}$ (0.35 Jy s), $4.2 \times 10^{-11}$ erg cm$^{-2}$ (4.8 Jy $\cdot$ 10 ms), and $1.6 \times 10^{-11}$ erg cm$^{-2}$ (17.9 Jy ms), orders of magnitude deeper than any previous simultaneous optical limit on a magnetar burst. The IFI+Iqueye measurement can also place a more stringent constraint to the spectral index of the optical to hard X-ray fluence of SGR J1935+2154, implying a spectrum steeper than $ν^{0.64}$. Fast optical timing observations of bursts associated with radio emission have then the potential to yield a detection.

astro-ph.HE↗

arXiv2021

Stellar intensity interferometry of Vega in photon counting mode

Stellar Intensity Interferometry is a technique based on the measurement of the second order spatial correlation of the light emitted from a star. The physical information provided by these measurements is the angular size and structure of the emitting source. A worldwide effort is presently under way to implement stellar intensity interferometry on telescopes separated by long baselines and on future arrays of Cherenkov telescopes. We describe an experiment of this type, realized at the Asiago Observatory (Italy), in which we performed for the first time measurements of the correlation counting photon coincidences in post-processing by means of a single photon software correlator and exploiting entirely the quantum properties of the light emitted from a star. We successfully detected the temporal correlation of Vega at zero baseline and performed a measurement of the correlation on a projected baseline of $\sim$2 km. The average discrete degree of coherence at zero baseline for Vega is $< g^{(2)} > \, = 1.0034 \pm 0.0008$, providing a detection with a signal-to-noise ratio $S/N \gtrsim 4$. No correlation is detected over the km baseline. The measurements are consistent with the expected degree of spatial coherence for a source with the 3.3 mas angular diameter of Vega. The experience gained with the Asiago experiment will serve for future implementations of stellar intensity interferometry on long-baseline arrays of Cherenkov telescopes.

astro-ph.IM↗

arXiv2020

Spin-down rate of the transitional millisecond pulsar PSR J1023+0038 in the optical band with Aqueye+

We present a timing analysis of the transitional millisecond pulsar PSR J1023+0038 using observations taken between January 2018 and January 2020 with the high time resolution photon counter Aqueye+ mounted at the 1.82 m Copernicus telescope in Asiago. We report the first measurement of the timing solution and the frequency derivative of PSR J1023+0038 based entirely on optical data. The spin-down rate of the pulsar is $(-2.53 \pm 0.04) \times 10^{-15}$ Hz$^2$, which is $\sim$20% slower than that measured from the X-ray observations taken in 2013-2016 and $\sim$5% faster than that measured in the radio band during the rotation-powered state.

astro-ph.HE↗

arXiv2019

The Cherenkov Telescope Array view of the Galactic Center region

Among all the high-energy environments of our Galaxy, the Galactic Center (GC) region is definitely the richest. It harbors a large amount of non-thermal emitters, including the closest supermassive black hole, dense molecular clouds, regions with strong star forming activity, multiple supernova remnants and pulsar wind nebulae, arc-like radio structures, as well as the base of what may be large-scale Galactic outflows, possibly related to the Fermi Bubbles. It also contains a strong diffuse TeV gamma-ray emission along the Galactic ridge, with a disputed origin, including the presence of a possible Pevatron, unresolved sources, and an increased relevance of the diffuse sea of cosmic rays. This very rich region will be one of the key targets for the next generation ground-based observatory for gamma-ray astronomy, the Cherenkov Telescope Array (CTA). Here we review the CTA science case for the study of the GC region, and present the planned survey strategy. These observations are simulated and we assess CTA's potential to better characterize the origin and nature of a selection of gamma-ray sources in the region.

astro-ph.HE↗

arXiv2019

Precise optical timing of PSR J1023+0038, the first millisecond pulsar detected with Aqueye+ in Asiago

We report the first detection of an optical millisecond pulsar with the fast photon counter Aqueye+ in Asiago. This is an independent confirmation of the detection of millisecond pulsations from PSR J1023+0038 obtained with SiFAP at the Telescopio Nazionale Galileo. We observed the transitional millisecond pulsar PSR J1023+0038 with Aqueye+ mounted at the Copernicus telescope in January 2018. Highly significant pulsations were detected. The rotational period is in agreement with the value extrapolated from the X-ray ephemeris, while the time of passage at the ascending node is shifted by $11.55 \pm 0.08$ s from the value predicted using the orbital period from the X-rays. An independent optical timing solution is derived over a baseline of a few days, that has an accuracy of $\sim 0.007$ in pulse phase ($\sim 12$ $μ$s in time). This level of precision is needed to derive an accurate coherent timing solution for the pulsar and to search for possible phase shifts between the optical and X-ray pulses using future simultaneous X-ray and optical observations.

astro-ph.HE↗