Search

Drowsy driving is a leading cause of fatal traffic accidents worldwide. Drowsy driving has emerged from modern societal trends such as long working hours, heavy reliance on vehicles, and insufficient sleep. Despite considerable efforts by researchers to develop efficient driver drowsiness detection systems, none so far has been widely adopted due to their high cost, intrusive nature, and ineffectiveness in challenging real-life situations. This paper presents a novel, real-time, non-intrusive, and cost-effective driver drowsiness detection system leveraging vision transformers (ViT). Our approach detects the driver’s face from each video frame and classifies the driver’s state as either ‘drowsy’ or ‘alert’ based on the entire facial image, as opposed to previous systems that rely on analyzing specific facial features. We demonstrate that the proposed Vision Transformers-based Driver Drowsiness Detection (ViT-DDD) system surpasses existing state-of-the-art methods, particularly in challenging scenarios such as drivers wearing glasses or sunglasses, or in different lighting conditions. The model was trained and evaluated on two widely used public drowsiness detection datasets, achieving classification accuracies of 98.89% on the NTHU-DDD dataset and 99.4% on the UTA-RLDD dataset. Furthermore, the system was successfully deployed on a Raspberry Pi microcomputer, integrated with an infrared camera, a GSM/GPS module, and a buzzer to alert the driver and report the drowsiness condition to the vehicle owner. Testing the prototype yielded highly promising results, with the system’s strong performance attributed to the ViT-DDD system and advanced hardware. The promising test results suggest the potential of this system in significantly reducing accidents caused by drowsy driving, with future work aiming to expand its capabilities and integration into broader vehicular systems.

This paper presents a very compact, wideband, and enhanced-gain antenna for 5G applications. A simple single-layer millimeter wave (mm-wave) metamaterial lens (meta-lens) is used to improve the gain, aperture efficiency, and gain bandwidth of a slotted-patch antenna over a wide range of frequencies from 25 GHz to 31 GHz. The lens exhibits a metamaterial negative refractive index behavior, which is attributed to a substantial gain enhancement of around 4–5 dBi over the whole band compared to the gain values of the slotted patch antenna alone. The lens’s unit cell comprises a simple single-layer split ring resonator (SRR) whose dimensions are carefully chosen to improve transmitted power and suppress absorbed and reflected power. The meta-lens consists of $8\times 8$ subwavelength SRR unit cells. Each cell has an area of $1.6\times 1.6$ mm2, it is located in the near-field region closely above a slotted patch antenna to produce a total antenna size of $12.8\times 12.8\times 7.27$ mm3 ( $1.2~\lambda \times 1.2~\lambda \times 0.68~\lambda $ , where $\lambda $ is the free space wavelength at 28 GHz). The maximum gain of the proposed antenna is 12.7 dBi, the 1 dB gain bandwidth is 18%, the maximum aperture efficiency is 92%, and the −10 dB impedance bandwidth (10 dB B.W.) is 17%. This excellent combination of essential metrics is hard to realize at mm-wave using narrowband antenna structures (microstrip patch antennas), and the aperture efficiency is the highest thus far for such a class of antennas.

This article presents accurate, efficient and reliable small-signal model parameter extraction approaches applied to Gallium Nitride (GaN) High Electron Mobility Transistor (HEMT). Firstly, a scanning-based systematic model parameter extraction methodology is developed. Then, newly reported Optimization Algorithms (OAs) namely Marine Predators Algorithm (MPA), Pelican Optimization Algorithm (POA) and Tunicate Swarm Algorithm (TSA) in combination with direct extraction method are utilized to develop hybrid model parameter extraction methodologies. Lastly, both the scanning-based systematic and OA-based hybrid modelling procedures are thoroughly validated and demonstrated on a GaN HEMT grown on diamond substrate to identify their pros and cons in distinct application settings. Moreover, reliability, accuracy, convergence behavior, complexity and execution time of MPA-, POA- and TSA-based hybrid extraction procedures are also discussed. We found that both classes of the approaches are able to produce an excellent agreement between the measured and modelled S-parameters for a wide frequency range up to 40 GHz. However, OA-based hybrid modelling procedures are more physically relevant.

Artificial Neural Network (ANN) is frequently utilized for the development of behavioral models of Gallium Nitride (GaN) High Electron Mobility Transistors (HEMTs). However, exhaustive investigation concerning the ANN algorithms implemented in major programming platforms for small-signal behavioral models of GaN HEMTs is generally not available. To fill this void, this paper carefully examines and evaluates ANN algorithms implemented in MATLAB, Python and R software environments for the development of accurate and efficient GaN HEMTs modelling. At first, the ANN based models are developed using MATLAB, Python’s major frameworks namely Keras, PyTorch and Scikit-learn, and R’s ANN framework namely H2O to model the GaN devices. Thereafter, an in-depth analysis is carried out to comprehend the usefulness of each framework in different application scenarios. At last, a detailed evaluation of the developed models in terms of generalization capability, training and prediction speed, seamless integration with the standard circuit design tool advanced design system, and of the development environments in respect of support and documentation, user-friendly interface, ease of model development, open-access and cost is carried out.

In this paper, the impact of AlGaN barrier thickness ( $\text{t}_{\mathrm{ AlGaN}}$ ) and substrate leakage on the noise conductance and noise figure in GaN High Electron Mobility Transistor (HEMT) is investigated. The investigation and analysis in this paper are targeting the Low Noise Amplifier (LNA) applications. The noise analysis is carried out using Technology Computer-Aided Design (TCAD) physical simulator. Initially, the DC, RF, and noise simulations are validated against measurements of a GaN device. AlGaN barrier thickness ( $\text{t}_{\mathrm{ AlGaN}}$ ) is varied and its impact on the minimum noise figure (NFmin) is analyzed. It is observed that the NFmin decreases with $\text{t}_{\mathrm{ AlGaN}}$ reduction at the typical bias conditions of LNA. This observation on the impact of $\text{t}_{\mathrm{ AlGaN}}$ on the NFmin follows Fukui’s model, which states that the NFmin decreases with the increase in transconductance. In addition, the impact of the substrate material on noise performance is analyzed. The substrates used for the investigation are Silicon (Si) and Silicon Carbide (SiC). At 40 GHz, it is found that the noise conductance and the NFmin of GaN HEMT on SiC substrate is reduced by 13% and 12%, respectively, in comparison with GaN HEMT on Si substrate. This could be attributed to the lower gate-to-substrate capacitance of the GaN HEMT on SiC substrate.

Abstract Machine learning‐based efficient temperature‐dependent small‐signal modelling approaches for GaN high electron mobility transistors (HEMTs) are presented by the authors here. The first method is an artificial neural network (ANN)‐based and makes use of the well‐known multilayer perceptron (MLP) architecture whereas the second technique is developed using support vector regression (SVR). The models are trained on a large set of measurement data obtained from a 2‐mm GaN‐on‐silicon device operating under varying operating conditions (bias voltages and ambient temperatures) over a wide frequency range of 0.1 to 20 GHz. An excellent agreement is found between the measured and the simulated S‐parameters for both models over the entire frequency range. It is identified that the training process and prediction capability of ANN is superior to SVR. However, the SVR is more robust when compared to the artificial neural network (ANN) in term of its sensitivity to local minima and uniqueness of the final solution. Subsequently, the performances of the proposed ANN‐ and SVR‐based models are improved by incorporating particle swarm optimization (PSO) in the model development process. The PSO improves the uniqueness of the ANN model whereas it enhances the performance of the SVR by optimising its control parameters. The proposed models exhibit very good accuracy and scalability.

Abstract A reliable small‐signal modelling approach has been developed and applied on GaN‐on‐diamond high electron mobility transistor. The extrinsic elements' extraction procedure was improved to provide an accurate characterization for the quasistatic behaviour of the intrinsic transistor. The frequency independence of the intrinsic elements at active multibias condition has been considered as another objective in addition to measurements' fitting. Physical relevant values for the model elements have been obtained. The model accuracy was also validated by means of S‐parameters simulation, which showed a very good fitting of the measured data.

In this paper, a new modeling technique is proposed for extracting small-signal lumped-element equivalent-circuit models for microwave transistors. The proposed procedure is based on using an optimization approach that is improved by targeting a quasi-static behavior as additional objective function rather than only minimizing the error between the simulated and measured scattering parameters. The validity of the developed modeling methodology is successfully demonstrated by considering a 0.25x1000 $\mu \text{m}^{2}$ gallium nitride (GaN) high-electron-mobility transistor (HEMT) as a case study.

This article presents an extensive study and demonstration of efficient electrothermal large-signal GaN HEMT modeling approaches based on combined techniques of Genetic Algorithm (GA) with Artificial Neural Networks (ANN), and Particle Swarm optimization (PSO) with Support Vector Regression (SVR). Another promising Gaussian Process Regression (GPR) based large-signal modeling approach is also explored and presented. The GA-ANN addresses the typical problem of local minima associated with the backpropagation (BP) based ANN. The GA successfully aids in the determination of optimal initial values for BP-ANN and enables it to find a unique optimal solution after subsequent of iterations with higher rate of convergence. This is also achieved using PSO-SVR with lower optimization variables. The developed modeling techniques are demonstrated and used to simulate the gate and drain currents of a 2-mm GaN device. All the models are relatively simple, practical, and easy to implement. The gate and drain currents models are embedded in an equivalent large-signal circuit’s model and built in Advanced Design System (ADS) software. The implemented model is validated by large-signal measurements and very good fitting results have been obtained. The model also showed an accurate simulation for a nonlinear power amplifier with very good computational speed and convergence.

In this paper an improved Gray-Wolf-Optimization (GWO) based small-signal modeling is developed. The proposed method is demonstrated by modeling GaN High Electron Mobility Transistors (HEMTs) on SiC and Diamond substrates. The technique bases on engineering the optimization objective function to provide reliable values for the model parameters; while keeping a better fitting for the targeted measurements. The reliability of extraction has been further improved by using physical relevant condition to remove any unrealistic values during the optimization process. The modeling procedure was applied on 2x50- $\mu \text{m}$ , 8x150- $\mu \text{m}$ , 8x250- $\mu \text{m}$ and 16x250- $\mu \text{m}$ GaN HEMTs on SiC substrate in addition to 2x125- $\mu \text{m}$ and 4x125- $\mu \text{m}$ GaN HEMTs on Diamond substrate. Very good results were obtained for both technologies with an excellent fitting for the related measurements. The results also show the reliability of the developed technique and validate its applicability for small- and large-signal modeling applications.

This work presents a multi-fault classification system using artificial neural network (ANN) to distinguish between different faults in rotating machines automatically. Rotation frequency and statistical features, including mean, entropy, and kurtosis were considered in the proposed model. The effectiveness of this model lies in using Synthetic Minority Over-sampling Technique (SMOTE) to overcome the problem of imbalance data classes. Furthermore, the Relief feature selection method was used to find the most influencing features and thus improve the performance of the model. Machinery Fault Database (MAFAULDA) was deployed to evaluate the performance of the prediction models, achieving an accuracy of 97.1% which surpasses other literature that used the same database. Results indicate that handling imbalance classes hold a key role in increasing the overall accuracy and generalizability of multi-layer perceptron (MLP) classifier. Furthermore, results showed that considering only statistical features and rotational speed are good enough to get a model with high classification accuracy.

This paper presents an efficient artificial neural network (ANN) electrothermal modeling approach applied to GaN devices. The proposed method is based on decomposing the device nonlinearity into intrinsic trapping-induced and thermal-induced nonlinearities that can be simulated by low-order ANN models. The ANN models are then interconnected in the physics-relevant equivalent circuit to accurately simulate the transistor. Genetic algorithm (GA)-based training procedure has been implemented to find optimal values for the weights of the ANN models. The modeling approach is used to develop a large-signal model for a 1-mm gate-width GaN high-electron mobility transistor (HMET). The model has been implemented in the advanced design system (ADS) and it has been validated by pulsed and continues small- and large-signal measurements. The model simulations showed a very good agreement with the measurements and verify the validity of the developed technique for dynamic electrothermal modeling of active devices.