This Letter presents a dual linearly polarised microstrip antenna for base station applications. Firstly, a dual-polarised metasurface-based antenna element employing differential feeding technology is proposed. The proposed antenna element achieves a simulated bandwidth of 11.6% for |S ₁₁| < −15 dB, mutual coupling better than 30 dB, and the cross-polarisation discrimination level over 35 dB. Secondly, based on the antenna element, a 3 × 3 antenna array is eventually presented, which was fabricated and characterised. The measured bandwidth is 12% for |S ₁₁| < −10 dB, and the port-isolation is over 25 dB. The superiorities of low-profile and high port isolation indicate that the proposed design is a promising candidate for sub-6G base-station applications.

A compact and wideband slot antenna backed by a square substrate integrated waveguide (SIW) cavity is presented. By exciting the dual-slotted SIW cavity with an offset coaxial feed, a transverse electric (TE)₁₁₀ resonant mode along with two hybrid modes as combinations of TE₁₁₀ and TE₁₂₀ resonances can be generated within a close range. The triple resonances contribute to producing a wideband response while maintaining a compact antenna structure. An antenna prototype with a total size of 0.49λ ₀ × 0.49λ ₀ × 0.16λ ₀ is manufactured and experimentally tested. Measures results show that an operating band of about 33.0% is obtained. Meanwhile, it also exhibits stable unidirectional patterns and moderate gains over the whole band.

In this Letter, an asynchronous spike-driven processor based on gated recurrent neural network algorithm for electrocardiogram (ECG) cardiac arrhythmias detection has been designed. Based on the processor, the proposed ECG detection model, containing a many-to-many gated recurrent unit layer and a fully connected layer, can achieve a high classification overall accuracy of 97.8% using the MIT-BIH arrhythmia database. The processor was fabricated in 55-nm 1P6M CMOS technology. It integrates about 1.4 million logic gates, including 40 KB of on-chip memory within a die area. The test chip consumes 6.28 mW at real-time operation frequency (200 MHz) with 1.2 V core supply voltage and at 1 kHz.

Vein texture elements in images are sometimes degraded due to centrally symmetric blur, e.g. due to motion of hand or camera not being in focus, which significantly affects recognition systems performance. Multiscale local phase quantisation (MLPQ) can effectively improve system accuracy in the presence of blur by dividing vein images into non-overlapping blocks and extracting multiscale local phase information. However, MLPQ misses some important discriminant information at the intersections of different regions, which leaves room for improving the system performance. In the present work, a multiscale overlapping blocks local phase quantisation (MOLPQ) histogram algorithm is presented to divide the vein image into overlapping blocks and extract multiscale local phase information, which includes the discriminant information lost in MLPQ. MOLPQ is validated via comparison with state-of-the-art recognition algorithms on a normal hand vein database, artificial-blur databases, and a normal-blur database, and the accuracies on the normal hand vein and normal-blur databases are 99.57 and 98.07%, respectively. Additionally, MOLPQ outperforms other methods on all databases in terms of accuracy.

In the current image processing field, medical image segmentation needs a lot of groundtruths, and the process of making these groundtruths is time-consuming and laborious. Thus, a novel retinal vessel segmentation algorithm without groundtruth is proposed in this Letter. The hierarchical clustering algorithm is first used to binary classify vessel and non-vessel pixels. Then classification results based on DRIVE databases are used as pseudo groundtruths to train the neural networks and transfer learning is considered for subsequent processing. Next the trained network is used as the feature extraction tool, by calculating and comparing the difference of image features between the target domain data (DRIVE database) and the source domain data (STARE, CHASE DB1, and HRF databases) extracted from the network. The data required for training is expanded based on semi-supervised clustering in this image feature space, finally the deep neural network is further fine-tuned. Experiments on the publicly available fundus image dataset DRIVE demonstrate that the proposed method outperforms many other state-of-the-art weakly supervised and unsupervised methods.

A compact and robust physically unclonable function (PUF) for secure key generation is proposed. With the help of a cross-coupled amplifier, the leakage-current mismatch between the two transistors was digitised to produce a unique chip identification. The measurement results show that the proposed PUF has the best-in-class randomness and uniqueness, as well as the following features: (i) an ultra-small PUF cell with a minimum feature size of 159 F², being the lowest reported to date in CMOS, (ii) a low native BER (bit-instability) of 0.16% (0.89%) was measured at the golden condition, and (iii) an autocorrelation of 0.007 at 95% confidence was achieved.

A novel multi-stage pipelined extended-counting (EC) for 3D-stacked CMOS image sensors is presented, which combines the benefits of a pipelined ADC, a first-order ADC and the EC principle in order to simultaneously achieve a high resolution and a high frame rate for 3D-stacked imagers. By assigning each stage of the pipeline to a sub-column of pixels while choosing another input stage of the pipeline when a sub-column has been read out, discontinuous rolling shutter artefacts like chopped images can be eliminated and frame rates up to 340 fps can be achieved for imagers with 8 K resolution.

This Letter presents an optimised cascaded integrator comb (CIC) filter sharpening algorithm embedding a fifth-order finite impulse response filter, which improves the worst-case alias rejection and simultaneously compensates the passband droop. The sharpening filter is designed without multipliers, which reduces hardware complexity. The results of a fourth-order CIC filter design practice show that the proposed algorithm improves the worst-case alias rejection to 128 dB, meanwhile the passband distortion is only 0.015 dB.

This Letter proposes a single event double-upset (SEDU)-fully-tolerant latch, referred to as FBSET, mainly featuring four interlocked branch circuits implemented by stacking three PMOS and one NMOS transistors or three NMOS and one PMOS transistors to achieve low power dissipation. The latch exhibits up to 84.56% area-power-delay product saving compared with recently reported latches. Simulation results validate that the proposed latch is completely immune to SEDU.

In this Letter, step-controllable radio frequency (RF) micro-electro-mechanical systems (MEMS) tunable capacitor with ultra-wide tuning range using netted membrane structure has newly proposed, fabricated, and characterised for highly miniaturised and reconfigurable RF devices and systems applications. The proposed tunable capacitor could step-wise increase the capacitance by introducing a few slits on the movable top membrane with a netted structure. The fabricated tunable capacitor exhibited 2.38–20.04 pF of tuning range and 146 (off-state) and 28 (full contacted-state) of quality factors at 1 GHz, respectively.

As one of the most fundamental tasks in computer vision, semantic segmentation assigns per pixel prediction of object categories. Training a robust model for semantic segmentation is challenging since pixel-level annotations are expensive to obtain. To alleviate the burden of annotations, the authors propose a weakly-supervised framework for zero-shot semantic segmentation, which can segment images having target classes without any pixel-level labelled instances. Under the assumption that the accessibility to image-level annotations of target classes does not violate the principle of zero pixel-level label in zero-shot semantic segmentation, we utilised image-level annotations to improve the proposed model's ability to extract pixel-level features. Furthermore, unlike existing zero-shot semantic segmentation methods, which use semantic embeddings as class embeddings to transfer knowledge from source classes to target classes, we use image-level features as their class embeddings to transfer knowledge since the distribution of pixel-level features is more similar to the distribution of image-level features rather than the distribution of semantic embeddings. Experimental results on the PASCAL-VOC data set under different data splits demonstrate that the proposed model achieves promising results.

This work considers a discrete-time multi-server queue with batch arrivals of packets having a deterministic length. By a novel analysis, the packet delay distribution is computed with a very light computational workload.

In this Letter, the physical layer security for a novel reconfigurable intelligent surface (RIS)-assisted non-orthogonal multiple access (NOMA) system in a multi-user scenario is investigated, where the authors consider the worst case that the eavesdropper also utilises the advantage of the RISs. More specifically, they derive analytical results for the secrecy outage probability (SOP). From the numerical results, they observe that the use of RISs can improve the secrecy performance compared to traditional NOMA systems. However, for the worst case that the received signals at the eavesdropper come from the RISs and source, increasing the number of intelligent elements on the RIS has a negative impact on the secrecy performance. At high signal-to-noise ratios, the system's SOP tends to a constant. Finally, the secrecy performance can be improved through group selection.

This Letter presents a new design of balanced-to-unbalanced (B2U) filtering magic-T based on circular patch resonator. Without additional circuit, the prescribed in-phase and out-of-phase specifications of the magic-T are achieved by elaborately studying and exploring the electric field distributions of the degenerate resonant modes in the circular patch resonators. Meanwhile, the desired filtering responses are attained by virtue of the coupling extent between the two designed patch resonators in the back-to-back form. In the proposed design, a pair of TM₁₁ modes at first make up passband. Meanwhile, nice common-mode suppression can be achieved at the balanced ports, and high isolation is achieved at the single-end ports, respectively. For demonstration, a prototype B2U filtering magic-T working at 4.3 GHz with 3 dB fractional bandwidth of 16.3% is implemented. Measured results coincide well with simulated ones, verifying the proposed design concept.

In this Letter, a new design of dual-band patch filter is proposed with greatly improved bandwidth and miniaturised size based on one single modified circular patch resonator. In this design, a metallic via is introduced in the centre region of the patch to develop the modified resonant TM₀₁ mode (TM′₀₁ mode), while two types of slot perturbations are adopted to separate a pair of degenerate modes of TM₁₁ (TM_11A and TM_11B) and control another pair of degenerate modes of TM₂₁ (TM_21A and TM_21B), respectively. Thus, two passbands with one transmission zero (TZ) can be attained by elaborately exploring these resonant modes (the TM′₀₁ and TM_11A modes form up the lower passband while TM_11B and TM_11A modes constitute the upper passband, and TM_21B mode is responsible for the TZ). For verification, one prototype is implemented. Results indicate that the novel dual-band filter (centred at 3.7 and 5.3 GHz) exhibits not only relative compact size (0.36 λ _g × 0.36 λ _g), but also much larger 3-dB fractional bandwidths (27 and 13%), nice return loss (better than 21 dB) and insertion loss (better than 1.0 dB), compared with other reported patch filters.

The authors present a model-free policy-based reinforcement learning model that introduces perturbations on the pattern of a metasurface. The objective is to learn a policy that changes the size of the patches, and therefore the impedance in the sides of an artificially structured material. The proposed iterative model assigns the highest reward when the patch sizes allow the transmission along a constrained path and penalties when the patch sizes make the surface wave radiate to the sides of the metamaterial. After convergence, the proposed model learns an optimal patch pattern that achieves lateral confinement along the metasurface. Simulation results show that the proposed learned-pattern can effectively guide the electromagnetic wave through a metasurface, maintaining its instantaneous eigenstate when the homogeneity is perturbed. Moreover, the pattern learned to prevent reflections by changing the patch sizes adiabatically. The reflection coefficient shows that most of the power gets transferred from the source to the destination with the proposed design.

A broadband active metasurface is designed around quantum-cascade (QC) gain material for use in a broadband tunable QC vertical external-cavity surface-emitting laser (VECSEL). The metasurface is based on a unit cell containing two resonant waveguide elements that couple with the periodicity of the metasurface to produce a multi-resonant reflection spectrum. Simulated reflectance spectra exhibit full-width half-maximum bandwidths up to 50% of the centre frequency. The tested QC-VECSEL demonstrates up to 15 mW of peak pulse power at 77 K and supports multimode lasing from 2.85 to 3.9 THz (>30% fractional bandwidth around 3.37 THz). The frequency coverage is limited by how short the cavity can be made, rather than the metasurface bandwidth. Consistent multimoding results from non-uniform distribution of spectral energy across the surface.

A power factor correction (PFC) charge-pump (CP) bridgeless converter for LEDs is presented. In the input stage, coupled inductors, together with CP capacitors, are integrated into a half-bridge inverter, thus characterising the single stage. This topology even being bridgeless due to the positioning of the load operates with a bus voltage below the bending voltage. The CP configuration is used to reduce the LED's current ripple. The prototype achieves an output power of 60 W with high power factor, achieving the efficiency of up to 94%, 127 V_ac input, total harmonic distortion of input current, with harmonic magnitudes within the IEC 61000-3-2 standard class C the device limits and flicker in conformed with the recommendation of IEEE 1789.

In this Letter, considering the configuration of small antenna array aperture used for special usage, the authors proposed a target trajectory estimation method by unambiguous phase differences with prior knowledge on target velocity from a single fixed passive sensor. First, they developed the maximum likelihood estimator of the target trajectory parameters based on unambiguous phase differences and then deduce the corresponding Cramer–Rao lower bound. Furthermore, in terms of the error in prior knowledge on target velocity, the influence of target velocity mismatch on trajectory parameters estimation is analysed. Finally, they evaluated the performance of the proposed method and validated the correctness of their analysis via numerical simulation.

In this Letter, a 1.2 kV SiC power MOSFET with a partially widened retrograde P-well (RP) structure and N-implanting region is proposed to enhance the device's reliability. Compared with the conventional SiC power MOSFET, the short circuit (SC) ability of the proposed structure can be improved effectively without sacrificing other performance. Simulation results reveal that a 40% reduction of the SC saturation current can be achieved, resulting in the SC withstand time increase from 7 to 10 μs at the DC-link voltage 800 V. Moreover, a better gate oxide reliability at 1.2 kV blocking condition also can be achieved. The peak electric field at the gate oxide interface is decreased by ∼48% owing to the shielding effect of the RP structure. In addition, the fabrication technology of the proposed structure is compatible with the standard planar SiC MOSFET manufacture process only with a few additional implanting steps. Therefore, this new MOSFET structure provides a simple and effective way to optimise the reliability of the planar SiC power MOSFET.

Dual-channel thin-film transistors (TFTs) show excellent effective mobility (μ _eff) and reliability, including negative-bias-illumination stability. To enhance the performance of these devices, post-thermal annealing in air is performed for 2 h at the low temperature of 250°C. To investigate the effects of low-temperature thermal annealing, indium–zinc-oxide/gallium–indium–zinc-oxide (IZO/IGZO) dual-channel TFTs with various IZO thicknesses are fabricated and examined. The observed parameters are μ _eff, saturation mobility (μ _sat), subthreshold slope (SS), and threshold voltage (V _TH). The interface quality may be improved by low-temperature thermal annealing, which was confirmed by the improvement in μ _eff and SS. However, these performance enhancements are intensively observed only in the channels with an IZO thickness below 10 nm. When the IZO thickness becomes 10 nm or more, the improvement in μ _eff due to low-temperature thermal annealing hardly occurs. This is because μ _eff is greatly influenced by the interface quality due to Coulomb scattering when the IZO thickness is <10 nm. However, for an IZO thickness above 10 nm, as phonon scattering also has a significant effect on μ _eff, the improvement in interface quality by thermal annealing is small. This is also supported by low-frequency noise measurements, which are sensitive to interface quality.

Extreme learning machine (ELM) is a single layer feed-forward neural network with advantages of fast training and good generalization properties. However, when the size of the hidden layer is increased, both of these advantages are lost as the redundant information may cause overfitting. Traditional way to deal with the issue is to introduce regularisation which promote sparsity but in the output layer weight matrix. In this Letter, we proposed the use of sparsity inside the output of the hidden layer such as to use it as the only non-linearity in the hidden layer. In the proposed formulation, we use linear activation function inside the hidden layer and keep k highest activity-bearing neurons as a measure of sparsity. Using the principal component analysis, we project the resulting output layer matrix onto a low-dimensional space in order to further remove redundant and irrelevant information, and speed up the training process. In order to verify the feasibility and effectiveness of the proposed method, we test and compare it with a number of ELM variants using benchmark datasets. Compared with these methods, our results demonstrate that the proposed method achieves better accuracy performance consistently across many different benchmark datasets.

The time-of-arrival (TOA) estimation method is an important ranging and positioning technology for indoor localisation and navigation. In this Letter, a novel TOA estimation method for smartphones based on a built-in microphone sensor is proposed, which utilises the propagation delay between optical and acoustic signals to achieve distance estimation. Unlike other TOA estimation technologies based on multi-sensors, a single MEMS microphone of a smartphone is used as the receiving sensor for both optical and acoustic signals, which is beneficial for eliminating the synchronisation error and improving ranging precision. Experimental results demonstrate that the proposed TOA estimation method is effective and accurate with an average ranging error of <0.29 m under a signal-to-noise ratio (SNR) of 1.18 dB for the smartphones tested in this work.

This Letter proposes a stochastic model of the wearable sensor nodes of a wireless body area network (WBAN) based on Markov decision process (MDP). The proposed technique tackles the problem of maximising the likelihood of data transmission while minimising the computational cost and energy. Suitable transmission strategies subject to the energy level of the nodes and channel conditions are needed to be investigated to strike a balance between network lifetime and packet delivery ratio. MDP formulation of the problem is solved offline to find out the optimal transmission strategies for the representative input conditions. The results of this offline computation approach can directly map to optimal transmission policies for intra-WBAN communications. The experimental results demonstrate the robustness of the proposed scheme in handling both favourable and unfavourable operating conditions.