arXiv subjects

Luca Benini

Publications and source records attributed to Luca Benini.

At least 91 records · Page 5Linked to original sources

arXiv2025

CEReBrO: Compact Encoder for Representations of Brain Oscillations Using Efficient Alternating Attention

Electroencephalograph (EEG) is a crucial tool for studying brain activity. Recently, self-supervised learning methods leveraging large unlabeled datasets have emerged as a potential solution to the scarcity of widely available annotated EEG data. However, current methods suffer from at least one of the following limitations: i) sub-optimal EEG signal modeling, ii) model sizes in the hundreds of millions of trainable parameters, and iii) reliance on private datasets and/or inconsistent public benchmarks, hindering reproducibility. To address these challenges, we introduce a Compact Encoder for Representations of Brain Oscillations using alternating attention (CEReBrO), a new small EEG foundation model. Our tokenization scheme represents EEG signals at a per-channel patch granularity. We propose an alternating attention mechanism that jointly models intra-channel temporal dynamics and inter-channel spatial correlations, achieving 2x speed improvement with 6x less memory required compared to standard self-attention. We present several model sizes ranging from 3.6 million to 85 million parameters. Pre-trained on over 20,000 hours of publicly available scalp EEG recordings with diverse channel configurations, our models set new benchmarks in emotion detection and seizure detection tasks, with competitive performance in anomaly classification and gait prediction. This validates our models' effectiveness and efficiency.

cs.LG↗

arXiv2025

PhysioWave: A Multi-Scale Wavelet-Transformer for Physiological Signal Representation

Physiological signals are often corrupted by motion artifacts, baseline drift, and other low-SNR disturbances, which pose significant challenges for analysis. Additionally, these signals exhibit strong non-stationarity, with sharp peaks and abrupt changes that evolve continuously, making them difficult to represent using traditional time-domain or filtering methods. To address these issues, a novel wavelet-based approach for physiological signal analysis is presented, aiming to capture multi-scale time-frequency features in various physiological signals. Leveraging this technique, two large-scale pretrained models specific to EMG and ECG are introduced for the first time, achieving superior performance and setting new baselines in downstream tasks. Additionally, a unified multi-modal framework is constructed by integrating pretrained EEG model, where each modality is guided through its dedicated branch and fused via learnable weighted fusion. This design effectively addresses challenges such as low signal-to-noise ratio, high inter-subject variability, and device mismatch, outperforming existing methods on multi-modal tasks. The proposed wavelet-based architecture lays a solid foundation for analysis of diverse physiological signals, while the multi-modal design points to next-generation physiological signal processing with potential impact on wearable health monitoring, clinical diagnostics, and broader biomedical applications. Code and data are available at: github.com/ForeverBlue816/PhysioWave

cs.LG↗

arXiv2025

WaveFormer: A Lightweight Transformer Model for sEMG-based Gesture Recognition

Human-machine interaction, particularly in prosthetic and robotic control, has seen progress with gesture recognition via surface electromyographic (sEMG) signals.However, classifying similar gestures that produce nearly identical muscle signals remains a challenge, often reducing classification accuracy. Traditional deep learning models for sEMG gesture recognition are large and computationally expensive, limiting their deployment on resource-constrained embedded systems. In this work, we propose WaveFormer, a lightweight transformer-based architecture tailored for sEMG gesture recognition. Our model integrates time-domain and frequency-domain features through a novel learnable wavelet transform, enhancing feature extraction. In particular, the WaveletConv module, a multi-level wavelet decomposition layer with depthwise separable convolution, ensures both efficiency and compactness. With just 3.1 million parameters, WaveFormer achieves 95% classification accuracy on the EPN612 dataset, outperforming larger models. Furthermore, when profiled on a laptop equipped with an Intel CPU, INT8 quantization achieves real-time deployment with a 6.75 ms inference latency.

cs.CV↗

arXiv2025

FEMBA: Efficient and Scalable EEG Analysis with a Bidirectional Mamba Foundation Model

Accurate and efficient electroencephalography (EEG) analysis is essential for detecting seizures and artifacts in long-term monitoring, with applications spanning hospital diagnostics to wearable health devices. Robust EEG analytics have the potential to greatly improve patient care. However, traditional deep learning models, especially Transformer-based architectures, are hindered by their quadratic time and memory complexity, making them less suitable for resource-constrained environments. To address these challenges, we present FEMBA (Foundational EEG Mamba + Bidirectional Architecture), a novel self-supervised framework that establishes new efficiency benchmarks for EEG analysis through bidirectional state-space modeling. Unlike Transformer-based models, which incur quadratic time and memory complexity, FEMBA scales linearly with sequence length, enabling more scalable and efficient processing of extended EEG recordings. Trained on over 21,000 hours of unlabeled EEG and fine-tuned on three downstream tasks, FEMBA achieves competitive performance in comparison with transformer models, with significantly lower computational cost. Specifically, it reaches 81.82% balanced accuracy (0.8921 AUROC) on TUAB and 0.949 AUROC on TUAR, while a tiny 7.8M-parameter variant demonstrates viability for resource-constrained devices. These results pave the way for scalable, general-purpose EEG analytics in both clinical and highlight FEMBA as a promising candidate for wearable applications.

cs.LG↗

arXiv2025

Finetuning and Quantization of EEG-Based Foundational BioSignal Models on ECG and PPG Data for Blood Pressure Estimation

Blood pressure (BP) is a key indicator of cardiovascular health. As hypertension remains a global cause of morbidity and mortality, accurate, continuous, and non-invasive BP monitoring is therefore of paramount importance. Photoplethysmography (PPG) and electrocardiography (ECG) can potentially enable continuous BP monitoring, yet training accurate and robust machine learning (ML) models remains challenging due to variability in data quality and patient-specific factors. Recently, multiple research groups explored Electroencephalographic (EEG)--based foundation models and demonstrated their exceptional ability to learn rich temporal resolution. Considering the morphological similarities between different biosignals, the question arises of whether a model pre-trained on one modality can effectively be exploited to improve the accuracy of a different signal type. In this work, we take an initial step towards generalized biosignal foundation models by investigating whether model representations learned from abundant EEG data can effectively be transferred to ECG/PPG data solely with fine-tuning, without the need for large-scale additional pre-training, for the BP estimation task. Evaluations on the MIMIC-III and VitalDB datasets demonstrate that our approach achieves near state-of-the-art accuracy for diastolic BP (mean absolute error of 1.57 mmHg) and surpasses by 1.5x the accuracy of prior works for systolic BP (mean absolute error 2.72 mmHg). Additionally, we perform dynamic INT8 quantization, reducing the smallest model size by over 3.5x (from 13.73 MB down to 3.83 MB) while preserving performance, thereby enabling unobtrusive, real-time BP monitoring on resource-constrained wearable devices.

eess.SP↗

arXiv2025

Wearable and Ultra-Low-Power Fusion of EMG and A-Mode US for Hand-Wrist Kinematic Tracking

Hand gesture recognition based on biosignals has shown strong potential for developing intuitive human-machine interaction strategies that closely mimic natural human behavior. In particular, sensor fusion approaches have gained attention for combining complementary information and overcoming the limitations of individual sensing modalities, thereby enabling more robust and reliable systems. Among them, the fusion of surface electromyography (EMG) and A-mode ultrasound (US) is very promising. However, prior solutions rely on power-hungry platforms unsuitable for multi-day use and are limited to discrete gesture classification. In this work, we present an ultra-low-power (sub-50 mW) system for concurrent acquisition of 8-channel EMG and 4-channel A-mode US signals, integrating two state-of-the-art platforms into fully wearable, dry-contact armbands. We propose a framework for continuous tracking of 23 degrees of freedom (DoFs), 20 for the hand and 3 for the wrist, using a kinematic glove for ground-truth labeling. Our method employs lightweight encoder-decoder architectures with multi-task learning to simultaneously estimate hand and wrist joint angles. Experimental results under realistic sensor repositioning conditions demonstrate that EMG-US fusion achieves a root mean squared error of $10.6^\circ\pm2.0^\circ$, compared to $12.0^\circ\pm1^\circ$ for EMG and $13.1^\circ\pm2.6^\circ$ for US, and a R$^2$ score of $0.61\pm0.1$, with $0.54\pm0.03$ for EMG and $0.38\pm0.20$ for US.

eess.SP↗

arXiv2025

A Direct Memory Access Controller (DMAC) for Irregular Data Transfers on RISC-V Linux Systems

With the ever-growing heterogeneity in computing systems, driven by modern machine learning applications, pressure is increasing on memory systems to handle arbitrary and more demanding transfers efficiently. Descriptor-based direct memory access controllers (DMACs) allow such transfers to be executed by decoupling memory transfers from processing units. Classical descriptor-based DMACs are inefficient when handling arbitrary transfers of small unit sizes. Excessive descriptor size and the serialized nature of processing descriptors employed by the DMAC lead to large static overheads when setting up transfers. To tackle this inefficiency, we propose a descriptor-based DMAC optimized to efficiently handle arbitrary transfers of small unit sizes. We implement a lightweight descriptor format in an AXI4-based DMAC. We further increase performance by implementing a low-overhead speculative descriptor prefetching scheme without additional latency penalties in the case of a misprediction. Our DMAC is integrated into a 64-bit Linux-capable RISC-V SoC and emulated on a Kintex FPGA to evaluate its performance. Compared to an off-the-shelf descriptor-based DMAC IP, we achieve 1.66x less latency launching transfers, increase bus utilization up to 2.5x in an ideal memory system with 64-byte-length transfers while requiring 11% fewer lookup tables, 23% fewer flip-flops, and no block RAMs. We can extend our lead in bus utilization to 3.6x with 64-byte-length transfers in deep memory systems. We synthesized our DMAC in GlobalFoundries' GF12LP+ node, achieving a clock frequency of over 1.44 GHz while occupying only 49.5 kGE.

cs.AR↗

arXiv2025

relOBI: A Reliable Low-latency Interconnect for Tightly-Coupled On-chip Communication

On-chip communication is a critical element of modern systems-on-chip (SoCs), allowing processor cores to interact with memory and peripherals. Interconnects require special care in radiation-heavy environments, as any soft error within the SoC interconnect is likely to cause a functional failure of the whole SoC. This work proposes relOBI, an extension to the Open Bus Interface (OBI) combining triple modular redundancy (TMR) for critical handshake signals with error correction codes (ECC) protection on other signals. Implementing and testing the reliable crossbar shows improved reliability to injected single faults from a vulnerability of 34.85 % to zero compared to the irredundant baseline, with an area increase of 2.6 $\times$. The area overhead is 1.8 $\times$ lower than that reported in the literature for fine-grained triplication and voting.

cs.AR↗

arXiv2025

Systematic Prevention of On-Core Timing Channels by Full Temporal Partitioning

Microarchitectural timing channels enable unwanted information flow across security boundaries, violating fundamental security assumptions. They leverage timing variations of several state-holding microarchitectural components and have been demonstrated across instruction set architectures and hardware implementations. Analogously to memory protection, Ge et al. have proposed time protection for preventing information leakage via timing channels. They also showed that time protection calls for hardware support. This work leverages the open and extensible RISC-V instruction set architecture (ISA) to introduce the temporal fence instruction fence.t, which provides the required mechanisms by clearing vulnerable microarchitectural state and guaranteeing a history-independent context-switch latency. We propose and discuss three different implementations of fence.t and implement them on an experimental version of the seL4 microkernel and CVA6, an open-source, in-order, application class, 64-bit RISC-V core. We find that a complete, systematic, ISA-supported erasure of all non-architectural core components is the most effective implementation while featuring a low implementation effort, a minimal performance overhead of less than 1%, and negligible hardware costs.

cs.CR↗

arXiv2025

Co-designing a Programmable RISC-V Accelerator for MPC-based Energy and Thermal Management of Many-Core HPC Processors

Managing energy and thermal profiles is critical for many-core HPC processors with hundreds of application-class processing elements (PEs). Advanced model predictive control (MPC) delivers state-of-the-art performance but requires solving an online optimization problem over a thousand times per second (1 kHz control bandwidth), with computational and memory demands scaling with PE count. Traditional MPC approaches execute the controller on the PEs, but operating system overheads create jitter and limit control bandwidth. Running MPC on dedicated on-chip controllers enables fast, deterministic control but raises concerns about area and power overhead. In this work, we tackle these challenges by proposing a hardware-software codesign of a lightweight MPC controller, based on an operator-splitting quadratic programming solver and an embedded multi-core RISC-V controller. Key innovations include pruning weak thermal couplings to reduce model memory and ahead-of-time scheduling for efficient parallel execution of sparse triangular systems arising from the optimization problem. The proposed controller achieves sub-millisecond latency when controlling 144 PEs at 500 MHz, delivering 33x lower latency and 7.9x higher energy efficiency than a single-core baseline. Operating within a compact less than 1 MiB memory footprint, it consumes as little as 325 mW while occupying less than 1.5% of a typical HPC processor's die area.

cs.DC↗

arXiv2025

Ultra-Efficient On-Device Object Detection on AI-Integrated Smart Glasses with TinyissimoYOLO

Smart glasses are rapidly gaining advanced functions thanks to cutting-edge computing technologies, especially accelerated hardware architectures, and tiny Artificial Intelligence (AI) algorithms. However, integrating AI into smart glasses featuring a small form factor and limited battery capacity remains challenging for a satisfactory user experience. To this end, this paper proposes the design of a smart glasses platform for always-on on-device object detection with an all-day battery lifetime. The proposed platform is based on GAP9, a novel multi-core RISC-V processor from Greenwaves Technologies. Additionally, a family of sub-million parameter TinyissimoYOLO networks are proposed. They are benchmarked on established datasets, capable of differentiating up to 80 classes on MS-COCO. Evaluations on the smart glasses prototype demonstrate TinyissimoYOLO's inference latency of only 17ms and consuming 1.59mJ energy per inference. An end-to-end latency of 56ms is achieved which is equivalent to 18 frames per seconds (FPS) with a total power consumption of 62.9mW. This ensures continuous system runtime of up to 9.3 hours on a 154mAh battery. These results outperform MCUNet (TinyNAS+TinyEngine), which runs a simpler task (image classification) at just 7.3 FPS, while the 18 FPS achieved in this paper even include image-capturing, network inference, and detection post-processing. The algorithm's code is released open with this paper and can be found here: https://github.com/ETH-PBL/TinyissimoYOLO

cs.CV↗

arXiv2025

Foundation Models for Structural Health Monitoring

Structural Health Monitoring (SHM) is a critical task for ensuring the safety and reliability of civil infrastructures, typically realized on bridges and viaducts by means of vibration monitoring. In this paper, we propose for the first time the use of Transformer neural networks, with a Masked Auto-Encoder architecture, as Foundation Models for SHM. We demonstrate the ability of these models to learn generalizable representations from multiple large datasets through self-supervised pre-training, which, coupled with task-specific fine-tuning, allows them to outperform state-of-the-art traditional methods on diverse tasks, including Anomaly Detection (AD) and Traffic Load Estimation (TLE). We then extensively explore model size versus accuracy trade-offs and experiment with Knowledge Distillation (KD) to improve the performance of smaller Transformers, enabling their embedding directly into the SHM edge nodes. We showcase the effectiveness of our foundation models using data from three operational viaducts. For AD, we achieve a near-perfect 99.9% accuracy with a monitoring time span of just 15 windows. In contrast, a state-of-the-art method based on Principal Component Analysis (PCA) obtains its first good result (95.03% accuracy), only considering 120 windows. On two different TLE tasks, our models obtain state-of-the-art performance on multiple evaluation metrics (R$^2$ score, MAE% and MSE%). On the first benchmark, we achieve an R$^2$ score of 0.97 and 0.90 for light and heavy vehicle traffic, respectively, while the best previous approach (a Random Forest) stops at 0.91 and 0.84. On the second one, we achieve an R$^2$ score of 0.54 versus the 0.51 of the best competitor method, a Long-Short Term Memory network.

cs.LG↗

arXiv2025

A Parallel Ultra-Low Power Silent Speech Interface based on a Wearable, Fully-dry EMG Neckband

We present a wearable, fully-dry, and ultra-low power EMG system for silent speech recognition, integrated into a textile neckband to enable comfortable, non-intrusive use. The system features 14 fully-differential EMG channels and is based on the BioGAP-Ultra platform for ultra-low power (22 mW) biosignal acquisition and wireless transmission. We evaluate its performance on eight speech commands under both vocalized and silent articulation, achieving average classification accuracies of 87$\pm$3% and 68$\pm$3% respectively, with a 5-fold CV approach. To mimic everyday-life conditions, we introduce session-to-session variability by repositioning the neckband between sessions, achieving leave-one-session-out accuracies of 64$\pm$18% and 54$\pm$7% for the vocalized and silent experiments, respectively. These results highlight the robustness of the proposed approach and the promise of energy-efficient silent-speech decoding.

eess.SY↗

arXiv2025

Optimal Brain Restoration for Joint Quantization and Sparsification of LLMs

Recent advances in Large Language Model (LLM) compression, such as quantization and pruning, have achieved notable success. However, as these techniques gradually approach their respective limits, relying on a single method for further compression has become increasingly challenging. In this work, we explore an alternative solution by combining quantization and sparsity. This joint approach, though promising, introduces new difficulties due to the inherently conflicting requirements on weight distributions: quantization favors compact ranges, while pruning benefits from high variance. To attack this problem, we propose Optimal Brain Restoration (OBR), a general and training-free framework that aligns pruning and quantization by error compensation between both. OBR minimizes performance degradation on downstream tasks by building on a second-order Hessian objective, which is then reformulated into a tractable problem through surrogate approximation and ultimately reaches a closed-form solution via group error compensation. Experiments show that OBR enables aggressive W4A4KV4 quantization with 50% sparsity on existing LLMs, and delivers up to 4.72x speedup and 6.4x memory reduction compared to the FP16-dense baseline.

cs.CL↗

arXiv2025

Efficient and Accurate Downfacing Visual Inertial Odometry

Visual Inertial Odometry (VIO) is a widely used computer vision method that determines an agent's movement through a camera and an IMU sensor. This paper presents an efficient and accurate VIO pipeline optimized for applications on micro- and nano-UAVs. The proposed design incorporates state-of-the-art feature detection and tracking methods (SuperPoint, PX4FLOW, ORB), all optimized and quantized for emerging RISC-V-based ultra-low-power parallel systems on chips (SoCs). Furthermore, by employing a rigid body motion model, the pipeline reduces estimation errors and achieves improved accuracy in planar motion scenarios. The pipeline's suitability for real-time VIO is assessed on an ultra-low-power SoC in terms of compute requirements and tracking accuracy after quantization. The pipeline, including the three feature tracking methods, was implemented on the SoC for real-world validation. This design bridges the gap between high-accuracy VIO pipelines that are traditionally run on computationally powerful systems and lightweight implementations suitable for microcontrollers. The optimized pipeline on the GAP9 low-power SoC demonstrates an average reduction in RMSE of up to a factor of 3.65x over the baseline pipeline when using the ORB feature tracker. The analysis of the computational complexity of the feature trackers further shows that PX4FLOW achieves on-par tracking accuracy with ORB at a lower runtime for movement speeds below 24 pixels/frame.

cs.CV↗

arXiv2025

A 410GFLOP/s, 64 RISC-V Cores, 204.8GBps Shared-Memory Cluster in 12nm FinFET with Systolic Execution Support for Efficient B5G/6G AI-Enhanced O-RAN

We present HeartStream, a 64-RV-core shared-L1-memory cluster (410 GFLOP/s peak performance and 204.8 GBps L1 bandwidth) for energy-efficient AI-enhanced O-RAN. The cores and cluster architecture are customized for baseband processing, supporting complex (16-bit real&imaginary) instructions: multiply&accumulate, division&square-root, SIMD instructions, and hardware-managed systolic queues, improving up to 1.89x the energy efficiency of key baseband kernels. At 800MHz@0.8V, HeartStream delivers up to 243GFLOP/s on complex-valued wireless workloads. Furthermore, the cores also support efficient AI processing on received data at up to 72 GOP/s. HeartStream is fully compatible with base station power and processing latency limits: it achieves leading-edge software-defined PUSCH efficiency (49.6GFLOP/s/W) and consumes just 0.68W (645MHz@0.65V), within the 4 ms end-to-end constraint for B5G/6G uplink.

cs.DC↗

arXiv2025

Wireless Low-Latency Synchronization for Body-Worn Multi-Node Systems in Sports

Biomechanical data acquisition in sports demands sub-millisecond synchronization across distributed body-worn sensor nodes. This study evaluates and characterizes the Enhanced ShockBurst (ESB) protocol from Nordic Semiconductor under controlled laboratory conditions for wireless, low-latency command broadcasting, enabling fast event updates in multi-node systems. Through systematic profiling of protocol parameters, including cyclic-redundancy-check modes, bitrate, transmission modes, and payload handling, we achieve a mean Device-to-Device (D2D) latency of 504.99 +- 96.89 us and a network-to-network core latency of 311.78 +- 96.90 us using a one-byte payload with retransmission optimization. This performance significantly outperforms Bluetooth Low Energy (BLE), which is constrained by a 7.5 ms connection interval, by providing deterministic, sub-millisecond synchronization suitable for high-frequency (500 Hz to 1000 Hz) biosignals. These results position ESB as a viable solution for time-critical, multi-node wearable systems in sports, enabling precise event alignment and reliable high-speed data fusion for advanced athlete monitoring and feedback applications.

eess.SY↗

arXiv2025

HiM2SAM: Enhancing SAM2 with Hierarchical Motion Estimation and Memory Optimization towards Long-term Tracking

This paper presents enhancements to the SAM2 framework for video object tracking task, addressing challenges such as occlusions, background clutter, and target reappearance. We introduce a hierarchical motion estimation strategy, combining lightweight linear prediction with selective non-linear refinement to improve tracking accuracy without requiring additional training. In addition, we optimize the memory bank by distinguishing long-term and short-term memory frames, enabling more reliable tracking under long-term occlusions and appearance changes. Experimental results show consistent improvements across different model scales. Our method achieves state-of-the-art performance on LaSOT and LaSOText with the large model, achieving 9.6% and 7.2% relative improvements in AUC over the original SAM2, and demonstrates even larger relative gains on smaller models, highlighting the effectiveness of our trainless, low-overhead improvements for boosting long-term tracking performance. The code is available at https://github.com/LouisFinner/HiM2SAM.

cs.CV↗