We have measured surface EMG during swallowing of liquid and saliva.

This work investigates subspace non-parametric models for the task of learning a set of acoustic units from unlabeled speech recordings. We constrain the base-measure of a Dirichlet-Process mixture with a phonetic subspace—estimated from other source languages—to build an \emph{educated prior}, thereby forcing the learned acoustic units to resemble phones of known source languages. Two types of models are proposed: (i) the Subspace HMM (SHMM) which assumes that the phonetic subspace is the same for every language, (ii) the Hierarchical-Subspace HMM (H-SHMM) which relaxes this assumption and allows to have a language-specific subspace estimated on the unlabeled target data. These models are applied on 3 languages: English, Yoruba and Mboshi and they are compared with various competitive acoustic units discovery baselines. Experimental results show that both subspace models outperform other systems in terms of clustering quality and segmentation accuracy. Moreover, we observe that the H-SHMM provides results superior to the SHMM supporting the idea that language-specific priors are preferable to language-agnostic priors for acoustic unit discovery.

This paper presents a 48 V–1 V merged-two-stage hybrid-switched-capacitor converter with a Linear Extendable Group Operated Point-of-Load (LEGO-PoL) architecture for ultra-high-current microprocessors, featuring 3-D stacked packaging and coupled inductors for miniaturized size and vertical power delivery. The architecture is highly modular and scalable. The switched capacitor circuits are connected in series on the input side to split the high input voltage into multiple stacked voltage domains. The multiphase buck circuits are connected in parallel to distribute the high output current into multiple parallel current paths. It leverages the advantages of switched capacitor circuits and multiphase buck circuits to achieve soft charging, current sharing, and voltage balancing. The inductors of the multiphase buck converters are used as current sources to soft-charge and soft-switch the switched-capacitor circuits, and the switched-capacitor circuits are utilized to ensure current sharing among the multiphase buck circuits. A 780 A vertical stacked CPU voltage regulator with a peak efficiency of 91.1% and a full load efficiency of 79.2% at an output voltage of 1 V with liquid cooling is built and tested. This is the first demonstration of a 48 V–1 V CPU voltage regulator to achieve over 1 A/mm2 current density and the first to achieve 1,000 W/in3 power density. It regulates output voltage between 0.8 V and 1.5 V through the entire 780 A current range.

A bidirectional single-phase AC to DC converter with high-frequency isolation finds a wide range of applications, including charging electric vehicles in the vehicle to grid applications. A conventional two-stage solution, where an AC-DC front-end rectifier is connected to an isolated DC-DC converter, suffers from poor efficiency due to hard switching of the AC-DC stage and poor reliability due to the existence of an electrolytic capacitance at the interstage DC link. A Dual Active Bridge (DAB) based single-stage AC-DC converter with a potential of bidirectional power flow can overcome the problems of a two stage solution. A rich literature exists in search for finding an efficient modulation strategy for this converter. This paper presents two constant switching frequency modulation strategies that support bidirectional power flow at any power factor utilizing all three degrees of freedom in modulation, also known as triple phase-shift modulation (TPS). One of the strategies minimizes RMS high-frequency transformer current over the line cycle, and the other one optimizes peak current. Hence, the conduction loss and the component stress over the entire line cycle are minimized. ZVS conditions are met for all high-frequency switching devices for the whole operating region, while the AC side converter is line frequency switched, incurring negligible switching loss. AC line current waveforms are of high quality and free from low-frequency harmonics. UPF operation is of importance for single-phase power conversion. All possible modes that the converter will experience over a line cycle for UPF operation are elaborated through detailed analysis. The proposed strategies are validated through experiment and simulation with 230 V, 50 Hz AC grid, 400 V DC, UPF, and output power of 1.2 kW.

Single-channel speech enhancement algorithms have seen great improvements over the past few years. Despite these improvements, they still lack the efficiency of the auditory system in extracting attended auditory information in the presence of competing speakers. Recently, it has been shown that the attended auditory information can be decoded from the brain activity of the listener. In this paper, we propose two novel deep learning methods referred to as the Brain Enhanced Speech Denoiser (BESD) and the U-shaped Brain Enhanced Speech Denoiser (U-BESD) respectively, that take advantage of this fact to denoise a multi-talker speech mixture. We use a Feature-wise Linear Modulation (FiLM) between the brain activity and the sound mixture, to better extract the features of the attended speaker to perform speech enhancement. We show, using electroencephalography (EEG) signals recorded from the listener, that U-BESD outperforms a current autoencoder approach in enhancing a speech mixture as well as a speech separation approach that uses brain activity. Moreover, we show that both BESD and U-BESD successfully extract the attended speaker without any prior information about this speaker. This makes both algorithms great candidates for realistic applications where no prior information about the attended speaker is available, such as hearing aids, cellphones, or noise cancelling headphones. All procedures were performed in accordance with the Declaration of Helsinki and were approved by the Ethics Committees of the School of Psychology at Trinity College Dublin, and the Health Sciences Faculty at Trinity College Dublin.

How people walk often reveals key insights into health, quality of life and independence. Here, we propose a smartphone-based gait monitoring system which is sensitive and accurate enough to measure temporal gait parameters during unsteady walking, differentiate between normal and impaired gait, and recognise changes in the impaired gait depending on the use of medication or walking aid.

In industry NLP application, our manually labeled data has a certain number of noise data. We present a simple method to find the noise data and remove them. We select the noise data whose human label is not contained in the top-K model’s predictions. The experiment result shows that our method works. For industry deep learning application, our method improve the text classification accuracy from 80.5% to 90.6% in dev dataset, and improve the human-evaluation accuracy from 83.2% to 90.1%.  The self-predict and self-drop method of this paper can not improve the accuracy to more than 95%, without human labeling again for the training dataset.

With the massive growth in wireless systems and the internet of things vision, low-cost and rapid antenna measurements in echoic lab environments have attracted significant research, commercial, and educational interest in the last decade. In this paper, we present a low-cost antenna positioning system based on open-source positioning hardware, for in-situ 3-dimensional radiation pattern measurements. The positioner is an off-the-shelf Arduino robotic arm and can perform azimuth and elevation measurements over a hemisphere using different rotation mechanisms. The positioning system is used in conjunction with a commercial Vector Network Analyzer (VNA) and wire λ/2 antennas for the second VNA port. The measured radiation patterns of a flexible microstrip patch antenna using the proposed setup are compared to 3D full-wave simulations as well as anechoic chamber measurements. The measurement setup, excluding the VNA, costs less than $300, and does not require calibrated reference antennas, rotary connectors, or specific PC peripherals. The open-source design library is publicly available at git.soton.ac.uk/mahm1m19/antennameasurement_system/.

We report the design, fabrication and quantitative performance analysis of a low-cost, flexible carbon nanotube (CNT) network-based deoxyribonucleic acid (DNA) sensor. These sensors comprise an array of ink-jet printed silver (Ag) electrodes on a transparent polyethylene terephthalate (PET) flexible substrate, where a CNT network acts as a sensing layer. The DNA hybridization is studied by immobilizing single-stranded DNA (ssDNA) probes on the CNT surface; these probes recognize their complementary DNA target. Further, we have carried out a quantitative performance analysis of the flexible CNT biosensors using the analytic hierarchy process (AHP). We have identified various influencing factors and sub-factors (performance indicators), and quantified the performance of the flexible CNT biosensors in different measured states (before bending, during bending and after bending). Additionally, the noise and other external factors contributing to the measured real signal have been quantified. The interpretation of the overall outcome will enable the improvement of the performance of flexible biosensors fabricated through large-scale manufacturing for possible commercialization.

Knee cartilage and bone segmentation is critical for physicians to analyze and diagnose articular damage and knee osteoarthritis (OA). Deep learning (DL) methods for medical image segmentation have largely outperformed traditional methods, but they often need large amounts of annotated data for model training, which is very costly and time-consuming for medical experts, especially on 3D images. In this paper, we report a new knee cartilage and bone segmentation framework, KCB-Net, for 3D MR images based on sparse annotation. KCB-Net selects a small subset of slices from 3D images for annotation, and seeks to bridge the performance gap between sparse annotation and full annotation. Specifically, it first identifies a subset of the most effective and representative slices with an unsupervised scheme; it then trains an ensemble model using the annotated slices; next, it self-trains the model using 3D images containing pseudo-labels generated by the ensemble method and improved by a bi-directional hierarchical earth mover’s distance (bi-HEMD) algorithm; finally, it fine-tunes the segmentation results using the primal-dual Internal Point Method (IPM). Experiments on two 3D MR knee joint datasets (the Iowa dataset and iMorphics dataset) show that our new framework outperforms state-of-the-art methods on full annotation, and yields high quality results even for annotation ratios as low as 5%.

Partial-structure-oriented work-energy theorem (WET) governing the work-energy transformation process of Yagi-Uda array antennas is derived. Driving power as the source to sustain a steady work-energy transformation is introduced. Employing WET and driving power, the essential difference between the working mechanisms of scattering objects and Yagi-Uda array antennas is revealed. The difference exposes that the conventional characteristic mode theory (CMT) for scattering objects cannot be directly applied to Yagi-Uda array antennas. Under WET framework, this paper proposes a generalized CMT for Yagi-Uda antennas. By orthogonalizing driving power operator (DPO), the WET-based CMT can construct a set of energy-decoupled characteristic modes (CMs) for an objective Yagi-Uda antenna, and then can provide an effective modal analysis for the Yagi-Uda antenna. In addition, a uniform interpretation for the physical meaning of the characteristic values / modal significances (MSs) of metallic, material, and metal-material composite Yagi-Uda antennas is also obtained by employing the WET-based modal decomposition and the field-current interaction expression of driving power.

Replacement of the exclusively designed instruments of the robotic surgery systems with the commercial hand-held wristed instruments provides advantages such as single-usability and cost reduction. A 4-DOF robotic system, based on a modified non-symmetric 2-DOF agile-eye mechanism, was developed to manipulate the hand-held wristed instruments. The kinematics of the mechanism was analyzed, its dimensions were optimized, and a functional prototype was tested experimentally. The optimized mechanism had a great kinematic isotropy (condition number <1.31) in the target workspace. Experimental studies revealed a high tracking accuracy ($0.27 +- 0.01 deg rms for the worse case) and a reasonably acceptable compliance (0.19 deg/N.m and 0.45 deg/N.m for the first and second kinematic chains respectively). By satisfying the design requirement, the robotic manipulator provides an attractive choice for robotic surgery systems. The performance of the manipulator can be improved further by increasing the stiffness of the second kinematic chain and performing kinematic calibration.

This paper proposes a Linear Parameter-Varying loop-shaping controller for a power-synchronized Phase-locked Loop (PLL)-less grid-following inverter (GFLI). This control strategy regulates the inverter’s output active and reactive powers at the terminal instead of the point of connection and does not require a PLL for extracting the voltage phase angle. Hence, the prevalent stability issues exhibited when GFLIs are connected to weak grids are not present, and the proposed GFLI control strategy can work under both very weak and strong grid conditions without being prone to instability. In this approach, the controller parameters are functions of the operating point and changed during the real-time operation such that the closed-loop performance is preserved in all operating points. Furthermore, since the grid impedance is a factor in the design process, a robustness analysis against grid impedance estimation error is conducted, and it is shown that discrepancies in estimated and real grid impedances are unlikely to make the system unstable. The performance of the proposed control design is validated in Matlab/PLECS and experiment for both strong and weak grids.

This paper considers security in eavesdropping attacks over wireless communication links in aeronautical telemetry systems. Here, we propose a secure coding technique for the integrated Network Enhanced Telemetry (iNET) communications system that can be coupled with modern encryption schemes. We consider a wiretap scenario where there are two telemetry links between a test article (TA) and a legitimate receiver, or ground station (GS). We show how these two links can be used to transmit both encrypted and unencrypted data streams while keeping both streams secure.

This literature review explains the construction and charge storage mechanisms in Lithium-ion batteries. Further, it elaborates on the electrode reactions in Lithium-ion batteries, and commonly used electrode materials and their structures. Different types of Lithium-based batteries' electrochemical performance were compared, in addition to other relevant differentiators. The energy storage mechanism in Supercapacitors is briefly touched upon -- and the electrochemical performance of supercapacitors is compared with that of lithium-ion batteries. Battery supercapacitor hybrids are introduced, with a brief section on their development over the past two decades following explanations of the charge storage mechanism and construction of battery supercapacitor hybrids. Battery supercapacitor hybrids are then compared with existing electrochemical energy storage mechanisms and finally, two types of battery supercapacitor hybrids were discussed.

Taking a cue from the Internet of Things, the Internet of Bodies (IoB) can be defined as a network of smart objects placed in, on, and around the human body, allowing for intra- and inter-body communications. This position paper aims to provide a glimpse into the opportunities created by implantable, injectable, ingestible, and wearable IoB devices. The paper starts with a thorough discussion of application-specific design goals, technical challenges, and enabling of communication standards. We discuss the reason that the highly radiative nature of radio frequency (RF) systems results in inefficient systems due to over-extended coverage that causes interference and becomes susceptible to eavesdropping. Body channel communication (BCC) presents an attractive, alternative wireless technology by inherently coupling signals to the human body, resulting in highly secure and efficient communications. The conductive nature of body tissues yields a better channel quality, while the BCC’s operational frequency range (1-100 kHz) eliminates the need for radio front-ends. State-of-the-art BCC transceivers can reach several tens of Mbps data rates at pJ/b energy efficiency levels that support IoB devices and applications. Furthermore, as the cyber and biological worlds meet, security risks and privacy concerns take center stage, leading to a discussion of the multi-faceted legal, societal, ethical, and political issues related to technology governance.

A fundamental challenge for machine learning models for electromagnetics is their ability to predict output quantities of interest (such as fields and scattering parameters) in geometries that the model has not been trained for. Addressing this challenge is a key to fulfilling one of the most appealing promises of machine learning for computational electromagnetics: the rapid solution of problems of interest just by processing the geometry and the sources involved. The impact of such models that can “generalize” to new geometries is more profound for large-scale computations, such as those encountered in wireless propagation scenarios. We present generalizable models for indoor propagation that can predict received signal strengths within new geometries, beyond those of the training set of the model, for transmitters and receivers of multiple positions, and for new frequencies. We show that a convolutional neural network can “learn” the physics of indoor radiowave propagation from ray-tracing solutions of a small set of training geometries, so that it can eventually deal with substantially different geometries. We emphasize the role of exploiting physical insights in the training of the network, by defining input parameters and cost functions that assist the network to efficiently learn basic and complex propagation mechanisms.

The environmental impact of connected and autonomous vehicles (CAVs) is still uncertain. Little is known about how CAVs operational behavior influences the environmental performance of network traffic, including conventional vehicles (CVs). In this paper, a microscopic traffic and emission modeling platform was applied to simulate CAVs operation in Motorway, Rural, and Urban road sections of a medium-sized European city, assuming different configurations of the car-following model parameters associated with a pre-determined or cooperative adaptative behavior of the CAVs. The main contribution is to evaluate the impact of the CAVs operation on the distribution of accelerations, Vehicle Specific Power (VSP) modal distribution, carbon dioxide (CO2) and nitrogen oxides (NOx) emissions for different road types and Market Penetration Rates (MPR). Results suggest CAVs operational behavior can affect CVs environmental performance either positively or negatively, depending on the driving settings and road type. It was found network-wide CO2 varies between savings of 18% and an increase of 4%, depending on the road type and MPR. CAVs adjusted driving settings allowed minimization of system NOx up to 13-23% for MPR ranging between 10 and 90%. These findings may support policymakers and traffic planners in developing strategies to better accommodate CAVs in a sustainable way.