In this work, a reference spur reduction method for integer-N subsampling PLLs is presented. A dual-loop architecture is implemented with the loops operating at opposite phases of the oscillator. A feedback network with error amplifier is there to equate the control voltages of both loops. Finally, the control voltages are added and applied to the oscillator, which is quite clean compared to that of the individual loops, aiding in spur reduction. The design is implemented in CMOS 90 nm technology with a oscillator frequency of 1 GHz. Reference spurs are at offsets of 100 MHz and post-layout simulation results show that the highest spurs are at -45dBc.
This paper proposes a computationally efficient polyphase implementation of analytically designed 2D FIR filters, with circular frequency response. The design is achieved in the frequency domain and is based on 1D prototype filters with low-pass, zero-phase, maximally-flat characteristic. The 2D FIR filter transfer function results directly in factored form. The filter implementation method is described in a simpler case, namely for a smaller size of the filter kernel and of the input image. Using decimation with small factors and the block filtering approach, we obtain a very efficient filter implementation at system level, with a low computational complexity.
People of all races and ages are prone to having acne especially in their teens and twenties. In particular, acne represents a disorder of the hair follicles and oil sebaceous glands. Clogged sebaceous glans with sebum which they secrete to moist skin, leads to pimples and cysts creation. Different types of acne lesions detection are important in diagnosis and treatment. This is why two different computer aided acne detection methods has been developed and extensively tested as presented in this paper. They are classified and identified by their primary characteristics, namely method A based on a saturation component involving sinus hyperbolics function and method B is focusing on contrast enhancement. As a result of testing of their specific applicability it was documented that the proposed methods show to be robust and offer a successful analysis tool in differentiating and detecting acne and scars areas in the images taken under various environmental and clinical conditions. While method A with its saturation component approach appears be more suitable in cases of subjects whose back show higher concentration of acne indicators, method B which uses contrast enhancement, offers advantage in cases of face focused acne indicators that are usually not uniform in affected populations.
The classical stereophonic acoustic echo cancellation (SAEC) setup requires the estimation of four acoustic impulse responses using four adaptive filters. The widely linear (WL) model was used in the past to simplify the manipulation of the SAEC scenarios, and allows the use of a single adaptive filter working with fewer complex valued signals and coefficients, requiring the same overall arithmetic resources. This paper compares the performances of the recursive least-squares (RLS) algorithm combined with several line search methods (LSMs) working within the WL framework. Simulation results are shown in order to demonstrate the performances of the proposed methods, and numerical estimations are presented regarding arithmetic complexities, with respect to practical applications and possible future developments.
This paper investigates the use of State Space design approach for the implementation of a controller dedicated to a multiphase buck converter. This approach benefits from using, in control, all phase currents along with the output voltage. While the proposed solution is generally applicable, the numerical example herein considers a 5-phase buck converter from 12 V input to 1.2 V output, under 100 A load. Design of the control system is done in MATLAB® following a pole placement design strategy. This paper investigates the practical (implementation) limitations of the method.
Super-resolution (SR) is the task of recovering High-Resolution (HR) images from given Low- Resolution (LR) Images. Various SR methods are available in the literature. The attention mechanism is one of the widely used approaches in the field of SR. In this paper, Counterfactual Attention Learning (CAL) based on causal inference is applied to increase the quality of attention in Face Super-Resolution (FSR). This approach helps to assess the quality of attention and provides a strong signal to supervise the learning activity. The paper discusses the effect of the learned attention on the task of SR through counterfactual intervention and the effect is maximized to make the model learn useful attention for FSR. The effectiveness of the method is tested, and upscaling is achieved using the Scale-Arbitrary SR model (ArbSR), which can handle both integer and non-integer scale factors. The experiments are carried out for different scale factors on the CelebA dataset. The results show that the technique enhances the performance of FSR task by both quality of the image and PSNR.
One of the attractive applications of nonlinear systems is noise generation, based on the chaotic behavior. The present contribution studies an analog hysteretic system exhibiting chaos and aiming at using the unpredictable generated signals for stochastic applications. Using different parameters for the state equations of the system reduces the possibility of identifying its behaviors. The statistical analysis presented complements the dynamic results better estimating the noise generation using hysteresis.
COVID-19 has been associated with several ECG abnormalities, hence this type of data has been investigated as a possible reliable indicator of the presence of the infection. The present paper aims to assess the performances of bag-of-words classifiers when processing digitized versions of paper multi-channel electrocardiograms. Four distinct feature extraction procedures and four different encoding strategies are evaluated in terms of the Area Under Curve (AUC), accuracy, sensitivity, and F1-score. Extensive experiments on an augmented version of a well-known public ECG dataset indicate that sparse encoding yields the best performances in line with more sophisticated classification approaches. The bag-of-words models are evaluated on various combinations of feature type, encoding strategy, and codebook dimension, showing significant robustness against all those parameters.
This article discusses simple methods for correcting FPN (Fixed Pattern Noise) in 1D linescan cameras. The correction methods vary in complexity and accuracy, with two of them requiring double and multipoint flat-field correction, while the third uses multi-slope regression. Provided are supporting experiments involving the illumination of a 1D sensor under various light levels. The results show that all correction methods significantly reduce the raw image FPN, but the multi-slope method provides the best linearity correction at lower illuminations with a relative error of less than 1%. Few conclusions were drawn on algorithm practicality in real-time applications with major trade-off being computational complexity.
Sensor positioning involves determining the best location for a sensor to be placed or installed so that it can effectively sense or measure the desired physical or environmental parameters. In this paper, we present a novel approach for finding sensor positions on a robotic hand using machine learning techniques. We focus on pressure sensors and their placement in order to enhance the performance and reliability of gesture recognition tasks. Our study analyzes data from 22 sensors placed at different locations on a right-handed robotic hand, simulating 10 distinct hand gestures. We employ various machine learning algorithms and create a correlation matrix to determine the most relevant sensor positions. The results highlight the significance of sensor optimization in improving the overall efficiency and effectiveness of robotic hand systems. By reducing the number of sensors to 12, we were still differentiate 10 hand gestures with % 99 accuracy.
The usage of internet services and online data transfers is increasing a lot nowadays. Security of network services should be a priority for every network service provider. It is important for any network owner to identify network vulnerabilities, especially if attackers try to access its services. It should detect intruders in a very short time and immediately block their actions in order to minimize the network damages and data loss. We design and simulate an intrusion detection platform based on virtual honeypots and present it in a given scenario. An algorithm with an increased capability of analyzing large volumes of data identifies cybersecurity risks by analyzing log files in order to make a decision regarding the appropriate countermeasures to take.
Calibrating an antenna is crucial to ensure an accurate and reliable measurement of electromagnetic fields. In this article, we propose a new method that combines the single-antenna method and the distance averaging method to calibrate an LPDA antenna. Our results show that the proposed method can provide a good calibration accuracy for a LPDA antenna, making it suitable for a wide range of applications in electromagnetic field measurement.
In this note we concentrate on the frequency-domain representation of linear time-varying systems (LTVS). Application of bilateral two-dimensional Laplace transform (2DLT) to LTV systems is invigorated. Specifically, transformation of the class of single-input single-output LTV systems is considered using the introduced 2D-to-lD reduction techniques.
Timely and accurate evaluation of land cover is essential in several planning and management activities. This study presents a novel framework for obtaining land cover clusters using the Sentinel-1 SAR and Sentinel-2 optical images in the Google Earth Engine (GEE) platform. It utilizes two new descriptors derived from Sentinel-1 and the conventional NDWI index obtained from Sentinel-2 data in the cloud-based GEE platform. The proposed methodology is evaluated for four cities from different continents globally. The clusters formed using the proposed framework were accurately compared with the World Cover 2020 (WC-20) map. We obtained an overall accuracy between 72% to 87% for the four cities.
Aiming at the difficulty of efficiency optimization control caused by unequal inductance parameters of interior permanent magnet synchronous motor(IPMSM), a PMSM efficiency optimization control strategy based on iterative algorithm is proposed. The equivalent transformation method is used to convert the total motor loss into functions related to the d axis current and find its partial derivative value. The optimal efficiency of the motor is optimized by using the partial derivative value and the Adam algorithm. This method can shorten the optimization time while achieving the best efficiency of the motor, and adjust the learning rate in real time to avoid the fluctuation of speed and torque caused by too fast current change. The experimental results show that compared with $i_{d}=0$ control and traditional loss model control, this method can effectively improve the motor efficiency and has fast dynamic response performance.
Irregularities of soil are defined by the term soil surface roughness and various factors affect it such as tillage operations, land management, soil texture, etc. Soil roughness impacts water infiltration and surface storage level, as well as wind and water erosion. We used two classical methods for soil roughness estimation based on chain and pinboard and tested their effectiveness in lab and in situ measurements. However, we concluded that even though these two methods are perfectly correlated when they are aligned on the same line over the sample surface, in-field results showed the opposite. Thus, we propose a new soil surface roughness measurement method based on fractal analysis of digital images of the soil surface, acquired using a camera obscura-based technique. We show that the 2D fractal analysis gives more pertinent results compared to the other methods designed for 1D measurements.
A one-layer frequency selective surface (FSS) is proposed for filtering applications. The initial structure consists of one square ring on one side of the supporting dielectric layer. By performing cylindrical bending on the initial structure, the stopband can be shifted, or multiple stopbands can be obtained. Parametric analyses of the frequency response as a function of substrate thickness and structure width are reported, demonstrating the flexibility of the design by allowing for the potential of modifying the stopbands. The analysis has been carried out by full-wave electromagnetic simulation in the sub-10 GHz frequency range.
Artificial intelligence (AI) models are popular nowadays due to their capacity to provide accurate classification results. A hardware implementation of an AI model offers major benefits such as deterministic and predictable functioning, low power consumption and the possibility of embedding the AI model in a mobile application. In this paper we present the hardware implementation of a Convolutional Neural Network (CNN) on a Xilinx Field Programmable Gate Array (FPGA) device. Our implementation is compact (small area) being aimed at very low power consumption. We describe the architecture and the choices we made for the parallel implementation. We prove the usefulness of our approach for simple trajectory classification. We conclude that our hardware implementation leads to interesting experimental results in terms of resource utilization, power consumption and running speed.
A nonlinear partial differential equation (PDE)- based photon-limited multi-channel image restoration technique is proposed here. It uses a novel well-posed vector-valued fourth-order reaction-diffusion model. The considered model leads to an anisotropic diffusion equation system whose single-valued PDEs are correlated by some shared terms. A finite difference-based discretization algorithm that solves numerically this reaction-diffusion system is then proposed. This fast-converging iterative numerical approximation scheme has been used successfully in the quantum denoising tests which are finally discussed in this paper.
In this paper we present a two steps method for palmprint classification, that combine texture feature and SURF descriptors extracted from palmprint images. Depending on the choice of parameters, the use of SURF descriptors can be a time-consuming procedure. Experiments based on the approach proposed in this paper have led to very good results, in a shorter computing time than SURF-based approaches. The experiments results demonstrate that the proposed method can be also used to select a subset of candidates in the classification process of an image, subset that can be further employed in other more computationally expensive methods. The method was tested on five well-known palmprint databases.
