Magnetic resonance imaging (MRI) relies on physiological triggering for cardiac gating, typically using the R-peak of an electrocardiogram (ECG). However, ECG-based triggering faces limitations in MRI environments due to challenges in electrode placement, magnetic field interference, and RF-induced heating. Alternatively, contact photoplethysmography (PPG) offers a feasible solution; however, it suffers from reduced accuracy and requires the fixation of a probe on the finger tip. To overcome these limitations, this paper proposes an MRI-compatible system for contactless forehead PPG using bundled fiber-optic guides. The proposed approach eliminates electrical interference and ensures safety. A low-power sensor node is proposed to investigate the trade-offs among signal fidelity, energy efficiency, and system latency in on-device PPG. By combining programmable optical sources, an analog front-end, and BLE connectivity, the platform enables reproducible MRI experimentation. It fully processes the PPG data in just 2.8 ms onboard, utilizing a low-power ARM Cortex-M33 core running at 128 MHz. A feasibility study involving 8 subjects was conducted to evaluate the proposed system and demonstrate the effectiveness of the MRI-compatible contactless PPG using green/red light. Several fiducial points of the PPG waveform - foot, onset, and peak - were evaluated for trigger generation. Despite a physiological delay of ∼100-150 ms relative to the R-peak of a reference ECG, the PPG-based R-peak point is reliably estimated with a jitter of 5.14 ms. The sensor node demonstrated the efficiency of the proposed solution operating with a 520 mAh battery for over 23 hours and integrates custom adapters for 2 m optical guides, ensuring safe electronic placement outside the MRI bore. These results confirm that our system can enable prospective gating of MRI measurements without the practical challenges of securing ECG leads or a fingertip PPG. The proposed system paves the way for a safe, contactless, low-power, self-contained sensor node with onboard processing and an interference-free gating method, with the potential to redefine physiological monitoring workflows in MRI environments, enabling precise synchronization without electromagnetic interference.

Achieving Simultaneous Localization and Mapping (SLAM) in an unfamiliar environment is a crucial challenge, especially for robots that rely on efficient on-device processing. While accurate mapping is achievable on high-end robotic systems, it still faces substantial challenges due to hardware and latency constraints, especially on smaller robots with limited power budget. Although machine learning is proving highly effective for robot perception, there is a growing need for lightweight solutions in terms of computation and sensing. This paper presents Nano VS, a lightweight monocular perception layer supporting semantic mapping with less than 1 M parameters. We propose a family of quantized and efficient models integrating emerging attention layers and weight-sharing in a multi-task neural network. Experimental results demonstrate multiple tasks within a single model, including Semantic Segmentation (SS), Feature Detection and Description (FDD), and Visual Place Recognition (VPR). Our findings indicate that multi-tasking effectively reduces computational overhead by eliminating the need for multiple networks. Nano VS achieves 70% classwise mIoU with the cityscapes benchmark and 66% Recall@1 in the Pitts30k challenge on tiny images (120x160 pixels). Finally, this paper implements and evaluates Nano VS on a novel milli-watt multi-core RISC-V Microcontroller (MCU), running the full semantic front-end in as little as 52 ms, consuming only 9 mJ per inference. This work represents a significant step towards making advanced SLAM capabilities accessible to tiny robots, or even faster and energy-efficient SLAM on high-end processors.

This work explores the feasibility of employing ultrasound (US) technology in a wrist-worn Internet-of-Things (IoT) device for low-power, high-fidelity heart rate (HR) extraction. US offers deep tissue penetration and can monitor pulsatile arterial blood flow in large vessels and the surrounding tissue, potentially improving robustness and accuracy compared to photoplethysmogram (PPG). We present an IoT wearable system prototype utilizing a commercial microcontroller (MCU) employing the onboard analogdigital converters (ADC) to capture high-frequency US signals and an innovative low-power US pulser. An envelope filter lowers the bandwidth of the US signal by a factor of >5 x, reducing the systems acquisition requirements without compromising accuracy (correlation coefficient between HR extracted from enveloped and raw signals, r(92)=0.996, p<0.001). The full signal processing pipeline is ported to fixed-point arithmetic for increased energy efficiency and runs entirely onboard. The extracted HR can be transmitted to the cloud via a Bluetooth low energy (BLE) module. The system has an average power consumption of 5.8mW, competitive with commercial PPG based systems, and the HR extraction algorithm requires only 69 kB of RAM and 71 ms of processing time on an ARM Cortex-M4 based MCU. The system is estimated to run continuously on a smartwatch battery for more than 7 days. To accurately evaluate the proposed circuit and algorithm and identify the anatomical location on the wrist with the highest accuracy for HR extraction, we collected a dataset from 10 healthy adults at three different wrist positions. The dataset comprises roughly 5 hours of HR data with an average of 80.6116.3 bpm. During recording, we synchronized the established electrocardiography (ECG) gold standard with our US-based method. The comparisons yield a Pearson correlation coefficient of r(92)=0.99, p<0.001 and a mean error of 0.6811.88 bpm in the lateral wrist position near the radial artery. Moreover, we tested our method while walking and running to assess its robustness to motion artifacts, achieving a heart rate extraction accuracy of 1.9912.80 bpm. The collected dataset and code used in this work have been open-sourced and are available at https://github.com/mgiordy/Ultrasound-Heart-Rate.

Fusion is a technique for merging multiple independently-trained neural networks in order to combine their capabilities. Past attempts have been restricted to the case of fully-connected, convolutional, and residual networks. This paper presents a systematic approach for fusing two or more transformer-based networks exploiting Optimal Transport to (soft-)align the various architectural components. We flesh out an abstraction for layer alignment, that can generalize to arbitrary architectures -- in principle -- and we apply this to the key ingredients of Transformers such as multi-head self-attention, layer-normalization, and residual connections, and we discuss how to handle them via various ablation studies. Furthermore, our method allows the fusion of models of different sizes (heterogeneous fusion), providing a new and efficient way to compress Transformers. The proposed approach is evaluated on both image classification tasks via Vision Transformer and natural language modeling tasks using BERT. Our approach consistently outperforms vanilla fusion, and, after a surprisingly short finetuning, also outperforms the individual converged parent models. In our analysis, we uncover intriguing insights about the significant role of soft alignment in the case of Transformers. Our results showcase the potential of fusing multiple Transformers, thus compounding their expertise, in the budding paradigm of model fusion and recombination. Code is available at https://github.com/graldij/transformer-fusion.

This paper introduces to a structured application of the One-Class approach and the One-Class-One-Network model for supervised classification tasks, specifically addressing a vowel phonemes classification case study within the Automatic Speech Recognition research field. Through pseudo-Neural Architecture Search and Hyper-Parameters Tuning experiments conducted with an informed grid-search methodology, we achieve classification accuracy comparable to nowadays complex architectures (90.0 - 93.7%). Despite its simplicity, our model prioritizes generalization of language context and distributed applicability, supported by relevant statistical and performance metrics. The experiments code is openly available at our GitHub.

In this demo we present a battery-free smart camera that exploits aggressive power management and energy harvesting to achieve face recognition in an energy-neutral fashion. A novel hardware accelerator for Convolution Neural Networks is employed to speed up the inference of the Tiny Machine Learning algorithm. The recognized face, and not the entire image, is sent via LoRa in a sensor network-like scenario. Experimental results demonstrated the capability of the developed sensor node to start and work perpetually with only a small photovoltaic panel array.