Hardware Limitations Research Articles

Deep Machine Learning of MobileNet, Efficient, and Inception Models

Today, specific convolution neural network (CNN) models assigned to specific tasks are often used. In this article, the authors explored three models: MobileNet, EfficientNetB0, and InceptionV3 combined. The authors were interested in investigating how quickly an artificial intelligence model can be taught with limited computer resources. Three types of training bases were investigated, starting with a simple base verifying five colours, then recognizing two different orthogonal elements, followed by more complex images from different families. This research aimed to demonstrate the capabilities of the models based on training base parameters such as the number of images and epoch types. Architectures proposed by the authors in these cases were chosen based on simulation studies conducted on a virtual machine with limited hardware parameters. The proposals present the advantages and disadvantages of the different models based on the TensorFlow and Keras libraries in the Jupiter environment based on the Python programming language. An artificial intelligence model with a combination of MobileNet, proposed by Siemens, and Efficient and Inception, selected by the authors, allows for further work to be conducted on image classification, but with limited computer resources for industrial implementation on a programmable logical controller (PLC). The study showed a 90% success rate, with a learning time of 180 s.

Quantum radio astronomy: Quantum linear solvers for redundant baseline calibration

The computational requirements of future large scale radio telescopes are expected to scale well beyond the capabilities of conventional digital resources. Current and planned telescopes are generally limited in their scientific potential by their ability to efficiently process the vast volumes of generated data. To mitigate this problem, we investigate the viability of emerging quantum computers for radio astronomy applications. In this a paper we demonstrate the potential use of variational quantum linear solvers in Noisy Intermediate Scale Quantum (NISQ) computers and combinatorial solvers in quantum annealers for a radio astronomy calibration pipeline. While we demonstrate that these approaches can lead to satisfying results when integrated in calibration pipelines, we show that current restrictions of quantum hardware limit their applicability and performance.

Multi-level feature extraction and reconstruction for 3D MRI image super-resolution

Magnetic resonance imaging (MRI) is an essential radiology technique in clinical diagnosis, but its spatial resolution may not suffice to meet the growing need for precise diagnosis due to hardware limitations and thicker slice thickness. Therefore, it is crucial to explore suitable methods to increase the resolution of MRI images. Recently, deep learning has yielded many impressive results in MRI image super-resolution (SR) reconstruction. However, current SR networks mainly use convolutions to extract relatively single image features, which may not be optimal for further enhancing the quality of image reconstruction. In this work, we propose a multi-level feature extraction and reconstruction (MFER) method to restore the degraded high-resolution details of MRI images. Specifically, to comprehensively extract different types of features, we design the triple-mixed convolution by leveraging the strengths and uniqueness of different filter operations. For the features of each level, we then apply deconvolutions to upsample them separately at the tail of the network, followed by the feature calibration of spatial and channel attention. Besides, we also use a soft cross-scale residual operation to improve the effectiveness of parameter optimization. Experiments on lesion-free and glioma datasets indicate that our method obtains superior quantitative performance and visual effects when compared with state-of-the-art MRI image SR methods.

Fire Fighting Robot with Human Detection and Audio Recognition

This paper presents innovative approaches to firefighting robots, addressing critical challenges in fire safety. CeaseFire, equipped with water and CO2 sprays, ensures real-time operations prioritizing human safety, offering user-friendly handling and extended control range. The DaNI(sbRIO)-powered robot excels in rapid fire detection and extinguishing, providing a reliable solution for hazardous situations. In parallel, the YOLOv4-tiny implemented on a limited hardware robot enhances accuracy in victim rescue during firefighting missions. The Arduino UNO-based mini fire-fighting robot autonomously detects and extinguishes fires, minimizing risks to firefighters and incorporating user intervention through an Android app. Additionally, a modular robot employs computer vision, temperature, and UV-TRON sensors for enhanced fire detection, pathfinding, and extinguishing, promising a significant advancement in autonomous firefighting technology. AFDER, an Autonomous Fire Detecting and Extinguishing Robot, integrates robotics, computer vision, and firefighting tech, autonomously navigating with lidar-based mapping and obstacle avoidance. With precise fire detection and targeted extinguishing capabilities, AFDER reduces human intervention, proving adaptable to diverse environments. Collectively, these advancements showcase the potential to revolutionize firefighting, safeguard lives, and address the complex challenges posed by fire emergencies.

Molecular Melodies: Unraveling the Hidden Harmonies of NMR Spectroscopy.

This work explores the evolution of auditory analysis in NMR spectroscopy, tracing its journey from a supplementary tool to visual methods such as oscilloscopes, to a technique sidelined due to technological advancements. Despite its renaissance in the late 1990s with artistic and scientific applications, widespread adoption was hindered by the necessity for hardware modifications and reliance on specialized software. Addressing these barriers, this paper introduces a new feature in Mnova NMR software that facilitates the easy auditory interpretation of NMR signals. We discuss new applications of this tool, emphasizing its utility in aiding the identification of specific functional groups by auditory analysis of the spectrum's multiplets, such as distinguishing between aromatic, olefinic, or aliphatic protons, thereby enriching the interpretative capabilities of NMR data.

High-resolution MRI synthesis using a data-driven framework with denoising diffusion probabilistic modeling

Objective. High-resolution magnetic resonance imaging (MRI) can enhance lesion diagnosis, prognosis, and delineation. However, gradient power and hardware limitations prohibit recording thin slices or sub-1 mm resolution. Furthermore, long scan time is not clinically acceptable. Conventional high-resolution images generated using statistical or analytical methods include the limitation of capturing complex, high-dimensional image data with intricate patterns and structures. This study aims to harness cutting-edge diffusion probabilistic deep learning techniques to create a framework for generating high-resolution MRI from low-resolution counterparts, improving the uncertainty of denoising diffusion probabilistic models (DDPM). Approach. DDPM includes two processes. The forward process employs a Markov chain to systematically introduce Gaussian noise to low-resolution MRI images. In the reverse process, a U-Net model is trained to denoise the forward process images and produce high-resolution images conditioned on the features of their low-resolution counterparts. The proposed framework was demonstrated using T2-weighted MRI images from institutional prostate patients and brain patients collected in the Brain Tumor Segmentation Challenge 2020 (BraTS2020). Main results. For the prostate dataset, the bicubic interpolation model (Bicubic), conditional generative-adversarial network (CGAN), and our proposed DDPM framework improved the noise quality measure from low-resolution images by 4.4%, 5.7%, and 12.8%, respectively. Our method enhanced the signal-to-noise ratios by 11.7%, surpassing Bicubic (9.8%) and CGAN (8.1%). In the BraTS2020 dataset, the proposed framework and Bicubic enhanced peak signal-to-noise ratio from resolution-degraded images by 9.1% and 5.8%. The multi-scale structural similarity indexes were 0.970 ± 0.019, 0.968 ± 0.022, and 0.967 ± 0.023 for the proposed method, CGAN, and Bicubic, respectively. Significance. This study explores a deep learning-based diffusion probabilistic framework for improving MR image resolution. Such a framework can be used to improve clinical workflow by obtaining high-resolution images without penalty of the long scan time. Future investigation will likely focus on prospectively testing the efficacy of this framework with different clinical indications.

Elderly-centered usability heuristics for augmented reality design and development

AbstractAugmented reality (AR) technology is making rapid progress and gaining high popularity, substantially due to the variety of new possibilities that it can offer to its users, including the elderly population. However, there is a lack of guidelines, recommendations, and testing techniques to make AR effective, efficient, and satisfying for its potential users. Usability heuristics are an effective tool that can help to ensure the usability of AR systems and can be used as a guide for the design and development process as well as a checklist for usability testing. However, general well-known usability heuristics do not consider specific aspects of AR, such as hardware limitations, privacy and safety issues, comfort of use, and the high importance of the user’s environment. General heuristics also do not always consider different user groups such as the elderly and people with disabilities. In this study, we developed a set of usability heuristics for AR systems for elderly users. The development process was inspired by the eight-step methodology by Quiñones et al. (Comput Stand Interfaces 59:109–129, 2018) and consisted of two main stages: gathering the heuristics from previous publications through an integrative literature review, and content analysis and validation of the collected heuristics through expert judgment (interviews with AR experts with industrial and academic backgrounds). The proposed final set of usability heuristics consists of six categories with a total of 55 heuristics that can be used by AR designers, developers, and usability testers to ensure the usability of AR systems for elderly users.

Beampattern analysis of cooperative beamforming with carrier frequency offsets in three-dimensional wireless networks

Cooperative beamforming (CB), where spatially distributed nodes synchronize their transmit phases to maximize the power of their combined signal at a desired receiver, can be used as an effective solution for hardware-limited nodes to achieve communication range extension. Thus, CB demonstrates effectiveness in providing higher reliability, lower latency, and extended transmission range for nodes in non-terrestrial networks (NTNs), which inherently face power and hardware limitations. In this paper, we consider the CB technique using UAVs in three-dimensional (3D) networks and analyze the average beampattern of the virtual antenna array constructed by the multiple UAVs. Further, because the mobile nature of the UAVs may cause carrier frequency offsets (CFOs), we analyze the impact of the CFO using the non-parametric kernel method. The simulation and analytical results show that the peak average beampattern degrades by about 3 dB with the CFO standard deviation of 1 kHz, which emphasizes the significance of frequency synchronization.

Subsurface Imaging Using Interferometry of Distributed Acoustic Sensing Ambient Noise Measurement along a Dark Fiber Line: A Case Study in Downtown Reno, Nevada

ABSTRACT Distributed acoustic sensing (DAS) technology is an emerging field of seismic sensing that enables recording ambient noise seismic data along the entire length of a fiber-optic cable at meter-scale resolution. Such a dense spatial resolution of recordings over long distances has not been possible using traditional methods because of limited hardware resources and logistical concerns in an urban environment. The low spatial resolution of traditional passive seismic acquisition techniques has limited the accuracy of the previously generated velocity profiles in many important urban regions, including the Reno-area basin, to the top 100 m of the underlying subsurface. Applying the method of seismic interferometry to ambient noise strain rate data obtained from a dark-fiber cable allows for generating noise cross correlations, which can be used to infer shallow and deep subsurface properties and basin geometry. We gathered DAS ambient noise seismic data for this study using a 12 km portion of a dark-fiber line in Reno, Nevada. We used gathered data to generate and invert dispersion curves to estimate the near-surface shear-wave velocity structure. Comparing the generated velocity profiles with previous regional studies shows good agreement in determining the average depth to bedrock and velocity variations in the analyzed domain. A synthetic experiment is also performed to verify the proposed framework further and better understand the effect of the infrastructural cover along the cable. The results obtained from this research provide insight into the application of DAS using dark-fiber lines in subsurface characterization in urban environments. It also discusses the potential effects of the conduit that covers such permanent fiber installations on the produced inversion results.

A dependable and resource-frugal ring oscillator physically unclonable function

ABSTRACT Silicon Physical Unclonable Function (PUF) is a crucial component for hardware security, providing unique device identification in networks. PUF provides robust security for cryptographic keys and IC digital fingerprints. However, resource constraints in network devices limit its use in security protocols. To address this challenge, PUF offers a lightweight hardware security solution with two main types: strong PUF (e.g. arbiter PUF) and Ring Oscillator (RO) PUF. The first option produces large CRPs with improved security but incurs hardware overhead, whereas the second option offers lower overhead with smaller CRP size, leading to reduced security. This paper introduces a novel, cost-effective solution to enhance RO PUF security while reducing hardware limitations. Our proposal involves a resource sharing technique with scan flip-flops, multiplexers, and inverter logic. This approach takes advantage of existing scan chains in circuit netlists to reduce PUF area without sacrificing uniqueness. A specific set of test patterns and odd number of inverters transform the memory-based PUF into an RO PUF. Our solution also leverages pass transistor-based implementation to significantly reduce power consumption and area overhead. Simulation results show that our approach effectively reduces area and power consumption while maintaining PUF uniqueness.

Learning dexterity from human hand motion in internet videos

To build general robotic agents that can operate in many environments, it is often useful for robots to collect experience in the real world. However, unguided experience collection is often not feasible due to safety, time, and hardware restrictions. We thus propose leveraging the next best thing as real world experience: videos of humans using their hands. To utilize these videos, we develop a method that retargets any 1st person or 3rd person video of human hands and arms into the robot hand and arm trajectories. While retargeting is a difficult problem, our key insight is to rely on only internet human hand video to train it. We use this method to present results in two areas: First, we build a system that enables any human to control a robot hand and arm, simply by demonstrating motions with their own hand. The robot observes the human operator via a single RGB camera and imitates their actions in real-time. This enables the robot to collect real-world experience safely using supervision. See these results at https://robotic-telekinesis.github.io. Second, we retarget in-the-wild human internet video into task-conditioned pseudo-robot trajectories to use as artificial robot experience. This learning algorithm leverages action priors from human hand actions, visual features from the images, and physical priors from dynamical systems to pretrain typical human behavior for a particular robot task. We show that by leveraging internet human hand experience, we need fewer robot demonstrations compared to many other methods. See these results at https://video-dex.github.io

Machine learning-based high-frequency neuronal spike reconstruction from low-frequency and low-sampling-rate recordings

Recording neuronal activity using multiple electrodes has been widely used to understand the functional mechanisms of the brain. Increasing the number of electrodes allows us to decode more variety of functionalities. However, handling massive amounts of multichannel electrophysiological data is still challenging due to the limited hardware resources and unavoidable thermal tissue damage. Here, we present machine learning (ML)-based reconstruction of high-frequency neuronal spikes from subsampled low-frequency band signals. Inspired by the equivalence between high-frequency restoration and super-resolution in image processing, we applied a transformer ML model to neuronal data recorded from both in vitro cultures and in vivo male mouse brains. Even with the x8 downsampled datasets, our trained model reasonably estimated high-frequency information of spiking activity, including spike timing, waveform, and network connectivity. With our ML-based data reduction applicable to existing multichannel recording hardware while achieving neuronal signals of broad bandwidths, we expect to enable more comprehensive analysis and control of brain functions.

ISP Agent: A Generalized In-storage-processing Workload Offloading Framework by Providing Multiple Optimization Opportunities

As solid-state drives (SSDs) with sufficient computing power have recently become the dominant devices in modern computer systems, in-storage processing (ISP), which processes data within the storage without transferring it to the host memory, is being utilized in various emerging applications. The main challenge of ISP is to deliver storage data to the offloaded workload. This is difficult because of the information gap between the host and storage, the data consistency problem between the host and offloaded workloads, and SSD-specific hardware limitations. Moreover, because the offloaded workloads use internal SSD resources, host I/O performance might be degraded due to resource conflicts. Although several ISP frameworks have been proposed, existing ISP approaches that do not deeply consider the internal SSD behavior are often insufficient to support efficient ISP workload offloading with high programmability. In this article, we propose an ISP agent, a lightweight ISP workload offloading framework for SSD devices. The ISP agent provides I/O and memory interfaces that allow users to run existing function codes on SSDs without major code modifications, and separates the resources for the offloaded workloads from the existing SSD firmware to minimize interference with host I/O processing. The ISP agent also provides further optimization opportunities for the offloaded workload by considering SSD architectures. We have implemented the ISP agent on the OpenSSD Cosmos+ board and evaluated its performance using synthetic benchmarks and a real-world ISP-assisted database checkpointing application. The experimental results demonstrate that the ISP agent enhances host application performance while increasing ISP programmability, and that the optimization opportunities provided by the ISP agent can significantly improve ISP-side performance without compromising host I/O processing.

A comparative study of cross-lingual sentiment analysis

This paper presents a detailed comparative study of the zero-shot cross-lingual sentiment analysis. Namely, we use modern multilingual Transformer-based models and linear transformations combined with CNN and LSTM neural networks. We evaluate their performance in Czech, French, and English. We aim to compare and assess the models’ ability to transfer knowledge across languages and discuss the trade-off between their performance and training/inference speed. We build strong monolingual baselines comparable with the current SotA approaches, achieving state-of-the-art results in Czech (96.0% accuracy) and French (97.6% accuracy). Next, we compare our results with the latest large language models (LLMs), i.e., Llama 2 and ChatGPT.We show that the large multilingual Transformer-based XLM-R model consistently outperforms all other cross-lingual approaches in zero-shot cross-lingual sentiment classification, surpassing them by at least 3%. Next, we show that the smaller Transformer-based models are comparable in performance to older but much faster methods with linear transformations. The best-performing model with linear transformation achieved an accuracy of 92.1% on the French dataset, compared to 90.3% received by the smaller XLM-R model. Notably, this performance is achieved with just approximately 0.01 of the training time required for the XLM-R model. It underscores the potential of linear transformations as a pragmatic alternative to resource-intensive and slower Transformer-based models in real-world applications. The LLMs achieved impressive results that are on par or better, at least by 1%–3%, but with additional hardware requirements and limitations. Overall, this study contributes to understanding cross-lingual sentiment analysis and provides valuable insights into the strengths and limitations of cross-lingual approaches for sentiment analysis.

Signals of opportunity for space navigation: An application-oriented review

Space navigation using X-Ray pulsars is known as opportunistic navigation, where we rely on signals of opportunity from natural phenomena to determine navigation solutions for spacecraft. However, this definition becomes too small or shallow, as it doesn't include other forms of navigation using natural beacons reported over the last two decades. Since a classification or review of all opportunistic techniques is still missing, this article presents the first survey on the field of signals of opportunity (SOOP), and introduces a framework for this scientific area, where natural beacons could be associated to stars, distant planets or natural phenomena, and signals could come with distinct characteristics and wavelengths. The new classification groups dozens of existing technologies of navigation using natural beacons, making easier for newcomers to interpret quickly a variety of works and mathematical formalisms. In addition, we deepen into an engineering perspective on the ongoing status of SOOP-based navigation, highlighting what is required to achieve the level of readiness on the methods that still on theory. We also give answer to a number of open questions, such as e.g. what criterion defines if a signal from a given natural beacon could be used or not for navigational purposes, as well as what is the minimum intersatellite distance that we can detect using an opportunistic signal, what type of signal characteristics we need from natural phenomena to generate navigation solutions or how far apart natural beacons preferentially need to be. These results enrich the field and become fundamental with an increasing interest of the space industry to bring novel forms of navigation into low-cost spacecraft of limited hardware capabilities (such as Cubesats).Furthermore, using results from past remote sensing missions, we preliminary found that Earth's lightning is a natural resource available in large scale and in sufficient variability that can be potentially used for relative opportunistic navigation. The physics of the problem is introduced, where at least theoretically, multiple lightning events can be used to map the arrival times to cooperating spacecraft, fixing their positions relative to each other independently from ground.

Effects of color cues on eye-hand coordination training with a mirror drawing task in virtual environment.

Mirror drawing is a motor learning task that is used to evaluate and improve eye-hand coordination of users and can be implemented in immersive Virtual Reality (VR) Head-Mounted Displays (HMDs) for training purposes. In this paper, we investigated the effect of color cues on user motor performance in a mirror-drawing task between Virtual Environment (VE) and Real World (RW), with three different colors. We conducted a 5-day user study with twelve participants. The results showed that the participants made fewer errors in RW compared to VR, except for pre-training, which indicated that hardware and software limitations have detrimental effects on the motor learning of the participants across different realities. Furthermore, participants made fewer errors with the colors close to green, which is usually associated with serenity, contentment, and relaxation. According to our findings, VR headsets can be used to evaluate participants' eye-hand coordination in mirror drawing tasks to evaluate the motor-learning of participants. VE and RW training applications could benefit from our findings in order to enhance their effectiveness.

Neutron-gamma discrimination with broaden the lower limit of energy threshold using BP neural network

Neutron-gamma discrimination is a tough and significative in experimental neutrons measurements procedure, especially for low-energy neutrons signal discrimination. In this work, based on the Pulse Shape Discrimination (PSD) and Back-Propagation (BP) artificial neural networks, a neutron-gamma discrimination method is developed to broaden the lower limit of energy threshold with the hidden layer of 20 neurons. Compared with neutron-gamma discrimination method based on PSD only, the developed neutron-gamma discrimination method based on the PSD and BP-ANN can discriminate neutron and gamma-ray signals with low energy threshold, which can discriminate signals up to 99.93%. Moreover, this work can reduce the energy threshold from 350 keV to 70 keV, as well as the acquired data utilization increased from 60% to more than 99.9%, which overcome the hardware limitations and distinguish neutron and gamma-ray signals, effectively. The developed neutron-gamma discrimination method and the trained neural network can be directly used to other experimental neutrons measurements.

A mini-review of signal processing techniques for RIS-assisted near field THz communication

Reflecting Intelligent Surfaces (RISs) are reshaping the landscape of wireless communications, particularly in the terahertz (THz) frequency bands, offering promising solutions to inherent challenges in the bands. THz communication boasts bandwidths exceeding 100 GHz, leading to data rates potentially in the terabits per second (Tbps) range, thereby making it an attractive proposition for wireless communications, imaging, and sensing. However, benefits come with challenges, including significant molecular absorption, scattering, diffraction, and hardware limitations. Moreover, as bandwidth in the THz range increases, so does the difficulty of signal processing at Nyquist rate. RIS emerges as a game-changer for 6G and beyond by providing programmable reflecting elements that can adaptively modify the phases and amplitudes of incident signals, enabling precision in directing THz waves and enhancing received signal strength. Such capabilities can significantly mitigate path loss and atmospheric absorption challenges. Furthermore, inherent pencil beamforming capabilities of RIS lead to optimized energy utilization. Major challenge in THz communications is the pressing needs for efficient algorithms for robust THz transceivers and optimizing RIS elements. This review describes the integration of RIS and near-field THz communications, highlighting their future potential and challenges for the next-generation wireless networks. In this article, a comprehensive understanding of the complexities and nuances of near-field propagation in 6G networks, especially as the technology shifts towards extremely large-scale antenna arrays (ELAA). Additionally, it will introduce the transformative potential of sub-Nyquist rate signal processing and artificial intelligence (AI) offering innovative solutions to address the inherent challenges of 6G communication, especially in channel estimation and beamforming strategies.

Single-image phase retrieval for off-the-shelf Zernike phase-contrast microscopes.

Quantitative phase imaging (QPI), such as digital holography, is considered a promising tool in the field of life science due to its noninvasive and quantitative visualization capabilities without the need for fluorescence labeling. However, the popularity of QPI systems is limited due to the cost and complexity of their hardware. In contrast, Zernike phase-contrast microscopy (ZPM) has been widely used in practical scenarios but has not been categorized as QPI, owing to halo and shade-off artifacts and the weak phase condition. Here, we present a single-image phase retrieval method for ZPM that addresses these issues without requiring hardware modifications. By employing a rigorous physical model of ZPM and a gradient descent algorithm for its inversion, we achieve single-shot QPI with an off-the-shelf ZPM system. Our approach is validated in simulations and experiments, demonstrating QPI of a polymer microbead and biological cells. The quantitative nature of our method for single-cell imaging is confirmed through comparisons with observations from an established QPI technique conducted through digital holography. This study paves the way for transforming non-QPI ZPM systems into QPI systems.

Edge-sorter: A hardware sorting engine for area & power constrained edge computing devices

In recent years, hardware sorters have been an attracted topic for researchers. Since hardware sorters play a crucial role in embedded systems, several attempts have been made to efficiently design and implement these sorters. Previous state-of-the-art hardware sorters are not suitable for embedded edge computing devices because they (1) consume high power, (2) occupy high area, (3) work for limited data-width numbers, (4) require many memory resources, and (5) finally, their architecture is not scalable with the number of input records. This paper proposes a hardware sorter for edge devices with limited hardware resources. The proposed hardware sorter, called Edge-Sorter, processes 4 bits of input records at each clock cycle. Edge-Sorter utilizes the unary processing in its main processing core. Edge-Sorter has valuable attributes compared to previous state-of-the-art techniques, including low power consumption, low area occupation, sorting numbers without storing their indices, sorting numbers with arbitrary data-width, and scalable with the number of input records. The proposed approach is evaluated and compared with previous state-of-the-art techniques with two different implementation and synthesis environments: Xilinx Vivado FPGA-based and Synopsys Design Compiler 45-nm ASIC-based. The Synthesis results of both environments indicate that both Edge-Sorter techniques reduces area and power consumption on average by 80% and 90%, respectively compared to previous techniques.