Feedforward Algorithm Research Articles

Structure design of fine-grained two-parallel Systolic FxLMS filter based on FPGA

In order to improve the feasibility of feedforward FxLMS algorithm in active noise reduction headphones and improve the convergence, throughput and power consumption of FxLMS filter, this paper proposes a fine-grained two-parallel Systolic FxLMS filter structure design scheme based on FPGA. Two-parallel fast FIR filters are used to improve the throughput of the system. Multi-operand addition, Multiplication-accumulation unit and Fusion multiplication-addition unit are introduced to compress the number of operations, so as to reduce the critical path delay. Experimental results show that the throughput of this structure is twice as high as that of the traditional structure at the same frequency, and the convergence is close to that of Systolic FxLMS algorithm; compared with the existing optimal structure, the clock speed of the proposed 8-tap filter is increased by 52.17%, and the power consumption is reduced by 9.28%.

NESM-γ: An Upper-Limb Exoskeleton With Compliant Actuators for Clinical Deployment

This letter describes the design and characterization of an upper-limb exoskeleton for post-stroke rehabilitation. The platform interacts with the shoulder and elbow of the user through four active joints, driven by series elastic actuators (SEAs) with custom springs to achieve compactness and ease of maintenance. The exoskeleton adopts a passive kinematic chain for aligning the user's and robot joints’ rotation axes, and a quick flipping mechanism to enable dual-side use. The pole-placement method based on the dynamic model of the SEA was used to design the low-level controller, to guarantee torque control precision and stability. The joint load due to the robot's gravity is counteracted by using a feed-forward gravity compensation algorithm. Experimental characterization demonstrates the torque control bandwidth up to 10 Hz and highly transparent behavior of the joints (namely, close to null parasitic impedance) at least up to 2 Hz, showing suitability for rehabilitation purposes.

Prediction of health-related quality of life through health literacy, perceived social support, spiritual well-being, and resilience in patients undergoing hemodialysis: An artificial neural network study

Background and aims: An important factor for determining the efficacy of treatment for patients under hemodialysis is the health-related quality of life (HRQOL). This study aimed to examine the predictor of HRQOL among hemodialysis (HD) patients based on spiritual well-being (SWB), perceived social support (PSS), health literacy (HL), and resilience. Methods: The present correlational study was performed on 260 patients randomly selected from the teaching hospitals of Ahvaz Jundishapur University of Medical Sciences. Data collection tools were a 36-item HRQQL questionnaire, Paloutzian & Ellison SWB scale, Zimet PSS scale, the Connor and Davidson resilience scale, and Montazeri and colleagues’ HL for Iranian adults. Python software was used to conduct data analysis. Descriptive statistics, feedforward, and backpropagation algorithm in an artificial neural network (ANN) was used to determine the effect of studied variables on HRQOL. Results: The mean age of patients was 51.73±15.32 years. Most of the patients were married (76.9%), had low-income level (55%), and had at least one comorbidity (59.2%). The mean duration of hemodialysis machine use was 5.06±3.43 years. Studied variables had grater impact on Physical QOL than the mental QOL. It was revealed that two dimensions of HL included evaluation skills (accuracy 0.92), accessibility skills (accuracy 0.92), and PSS (accuracy 0.88) were stronger predictors of physical QOL than other variables and SWB in the existence dimension (accuracy 0.80), understanding skills of HL (accuracy 0.76) and evaluation skills of HL (accuracy 0.75) were stronger predictors of mental QOL that other variables. Conclusion: To improve the HRQOL of HD patients, we should increase HL, support social relationships, and improve meaning in life by giving importance to the SWB of these patients in providing care services to them.

Resource optimization of MIMO using neural network for 5G communication

The principal objective of this work is to optimize the resources of massive MIMO techniques by an application of a neural network algorithm in order to reduce the manufacturing cost of 5G technology within the wireless peripheries of tropical regions. For plenty of annual precipitation in the tropical regions, the radio signals of these areas get attenuated in communicating any message i.e. signal fading occurs. Various diversity techniques have been used for diminishing signal fading in these tropical areas. MIMO technique is a spatial diversity technique used to minimize signal fading. In this paper, 1 × 1 to 256 × 256 MIMO techniques have been designed with the help of MATLAB and Xilinx system generator, and an 8 × 8 MIMO has been implemented in Xilinx Kintex7 FPGA. A CMOS logic technique has been incorporated to optimize the resources of different MIMO techniques, and for this purpose a two-layer feed forward neural network algorithm has been adopted. A comparison of the theoretical values obtained from the design by using both Xilinx system generator and Simulink with the values derived from the design by using CMOS logic gates, has been displayed in this paper. For neural network optimization, three types of training algorithms have been applied.

ANN modelling of surface roughness of FDM parts considering the effect of hidden layers, neurons, and process parameters

ABSTRACT Hidden layers perform a vital role in the performance of artificial neural networks (ANNs), especially for serpentine problems where the curtailment of accuracy and time complexity is observed. A higher number of hidden layers and neurons enlarge the order of weights and influence the ANN accuracy. In the present work, an ANN model is developed to predict the surface roughness of fused deposition modelling (FDM) parts considering the effect of the number of neurons, hidden layers, layer height, infill density, nozzle temperature, and print speed. A feedforward back-propagation machine learning algorithm is used to model an ANN. This study finds better prediction accuracy with the ANN architecture having two hidden layers with 150 neurons on each layer (R-squared: 0.875) followed by one hidden layer with 250 neurons (R-squared: 0.83). This study finds a decrease in the surface roughness of FDM parts with the increase in infill density. However, the surface roughness was observed to increase with the layer height, print speed, and nozzle temperature. This study finds a decrease in the surface roughness of FDM parts with the increase in infill density and an increase in the surface roughness with the layer height, print speed, and nozzle temperature.

High-Speed Silicon Integrated Polarization Stabilizer Assisted By a Polarimeter

Integrated polarization stabilizers along with corresponding stabilizing algorithms have been investigated for years to tackle the polarization fluctuation issue in polarization dependent systems such as photonics integrated circuits and coherent optical communication systems. Here, we propose and demonstrate a novel integrated polarization stabilizer which is composed of an in-line polarimeter and two cascaded Mach-Zehnder interferometers-based polarization converter, with a footprint of 2.8 mm 0.8 mm. Benefitting from the integrated in-line polarimeter and two controlling algorithms, namely feed-forward tracking algorithm and feedback tracking algorithm, the stabilizer can track the incoming light with arbitrary state of polarization in one-time manipulation. The whole process consumes about 30 s, which is approaching the limitation of the response time of the adopted thermal phase shifters, making the proposed device the fastest one with thermal phase shifters on silicon photonics platform. Meanwhile, the incoming light is endlessly tracked in real time and 99% converted to a horizontal linear polarized light.

An Empirical Study of Passengers’ Perceived Satisfaction with Monorail Service Quality: Case of Kuala Lumpur, Malaysia

The aim of the current study was to examine passengers’ perceived satisfaction with a monorail system by determining the factors that dominated their perception. For this study, 417 data points were collected through face-to-face questionnaire surveys in Kuala Lumpur, Malaysia, from 20 September 2019 to 10 December 2019. The study involved the use of several assessments to ascertain how the perceived satisfaction of passengers with the monorail service was influenced. The tests predicted which service factors fuelled this satisfaction and included an exploratory factor analysis, Spearman’s correlation test and an artificial neural network (ANN) model (termed the multilayer perceptron neural networks model) with a feed-forward backpropagation algorithm. The findings that were produced by these methods of analysing factors revealed the extraction of eight service-quality features as the key influences on the perceived satisfaction of passengers. The correlation test results revealed that these factors have a significant and positive relationship with perceived satisfaction. Finally, the ANN model with the optimum neuron number in the hidden layer is seven neurons. This model found that the dominant service quality factors that contributed to influencing passengers’ perceived satisfaction levels were (i) the provision of information, (ii) facilities, (iii) signage. The results of this study will benefit service providers, policymakers and planners in formulating effective strategies to enhance passengers’ satisfaction with the monorail service and increase the ridership.

Nonlinear Predictive Motion Control for Autonomous Mobile Robots Considering Active Fault-Tolerant Control and Regenerative Braking.

To further advance the performance and safety of autonomous mobile robots (AMRs), an integrated chassis control framework is proposed. In the longitudinal motion control module, a velocity-tracking controller was designed with the integrated feedforward and feedback control algorithm. Besides, the nonlinear model predictive control (NMPC) method was applied to the four-wheel steering (4WS) path-tracking controller design. To deal with the failure of key actuators, an active fault-tolerant control (AFTC) algorithm was designed by reallocating the driving or braking torques of the remaining normal actuators, and the weighted least squares (WLS) method was used for torque reallocation. The simulation results show that AMRs can advance driving stability and braking safety in the braking failure condition with the utilization of AFTC and recapture the braking energy during decelerations.

An excellent harmonic feedforward-sliding mode output feedback hybrid algorithm for helicopter active vibration control

Higher harmonic control algorithms in frequency domain are generally used for active vibration control of helicopter fuselage. However, these algorithms are inefficient to control transient vibration caused by rapid changing of helicopter flight conditions. Time domain feedback algorithms are efficient to control the transient vibration. In this paper, an excellent harmonic feedforward-sliding mode output feedback hybrid algorithm by combining frequency and time domain algorithms is presented to improve the effect and performance of active vibration control of helicopter fuselage, meanwhile to avoid the requirement of full state information of control system due to using the time domain algorithms. The convergence criterions of the harmonic feedforward algorithm and the hybrid algorithm are theoretically derived, respectively. The convergence speed of the hybrid algorithm is theoretically proved to be faster than that of the harmonic feedforward algorithm. Active vibration control simulation and experiment on a dynamically similar and geometrically scaled frame structure of a helicopter fuselage verify the proposed hybrid algorithm with excellent control effect and performance.

A flexible gripper with a wide-range variable stiffness structure based on shape memory alloy

PurposeVariable stiffness structure can significantly improve the interactive capabilities of grippers. Shape memory alloys have become a popular option for materials with variable stiffness structures. However, its variable stiffness range is limited by its stiffness in two phases. The purpose of this paper is to enhance the manipulation capabilities of tendon-driven flexible grippers by designing a wide-range variable stiffness structure.Design/methodology/approachConstitutive models of shape memory alloy and mechanical models are used to analyze the performance of the variable stiffness structure. A separated solution was used to combine the tendon-driven gripper and the variable stiffness structure. The feed-forward control algorithm is used to enhance the control stability of the variable stiffness structure.FindingsThe stiffness variable capability of the proposed variable stiffness structure is verified by experiments. The stability of the feedback control algorithm was verified by sinusoidal tracking experiments. The variable stiffness range of 8.41 times of the flexible gripper was tested experimentally. The interaction capability of the variable stiffness flexible gripper is verified by the object grasping experiments.Originality/valueA new wide-range variable stiffness structure is proposed and validated. The new variable stiffness structure has a larger range of stiffness variation and better control stability. The new flexible structure can be applied to conventional grippers to help them gain stiffness variable capability and improve their interaction ability.

ESTIMATION OF KNEE JOINT TORQUE DURING SIT–STAND MOVEMENT BASED ON sEMG SIGNALS USING NEURAL NETWORKS

The estimation of knee joint torque is important for the development of powered exoskeletons to achieve ideal gait characteristics. In this study, we proposed three different models to predict the required torque for performing sit-to-stand (STS) and back-to-sit (BTS) movements. The surface electromyography (sEMG) signals were extracted from the biceps femoris and rectus femoris muscles during STS and BTS movements. The time-domain features selected as input to the models for torque prediction are integrated EMG (iEMG), root mean square (RMS), and mean absolute value (MAV). Two-way ANOVA analysis identifies the significance of NN models and EMG features of the muscles in predicting the knee joint torque requirement. The artificial neural network models selected for prediction are the feed-forward back-propagation algorithm, ANFIS, and NARX. The theoretical value of knee joint torque calculated using the Lagrange method was compared with the torque output for each model based on root mean square error (RMSE). The desired torque predicted using the NARX model confirms to have the least average error ([Formula: see text][Formula: see text]Nm), which indicates that NARX can estimate knee joint torque more accurately from sEMG than other models.

A coalescent grid model of weighted mean temperature for China region based on feedforward neural network algorithm

A coalescent grid model of weighted mean temperature for China region based on feedforward neural network algorithm

Machine learning approach to predict the heat transfer coefficients pertaining to a radiant cooling system coupled with mixed and forced convection

Mixed convection phenomenon over radiant cooled surfaces with displacement ventilation in living environments is becoming a popular issue due to the airborne viruses and energy economy. Artificial neural networks are one of the machine learning methods that are widely evaluated as an engineering tool. In the current study, heat transfer coefficients for a radiant wall cooling system coupled with mixed and forced convection have been predicted by a machine learning approach. This approach should be noted as a first experimental investigation couple with an artificial neural network analysis in the open sources in which mixed convection systems in real sized living environments is examined. Experimentally obtained heat transfer coefficients have been used in the development of the feed forward back propagation multi-layer perceptron network structure. So as to analyze the impact of the input factors on the prediction performance, two neural network structures with dissimilar input parameters such as various temperatures, velocities, and heat transfer rates have been developed. By means of feed forward back propagation multi-layer perceptron neural network algorithms, convection, radiation, and total heat transfer coefficients have been predicted using the experimentally acquired dataset including 35 data points belonging to the mixed and forced convection conditions. Training, validation, and test data groups include 70%, 15%, and 15% of the dataset, in turn. Training algorithm has been computed via Levenberg-Marquardt one with 10 neurons in the hidden layer. The findings obtained from the computational solution have been evaluated as a result of the contrast with the target data with in the ±5% deviation band for all heat transfer coefficients. The performance factors have been computed and the estimation precision of the numerical models has been thoroughly examined.

Regression-based neural network for improving image reconstruction in diffuse optical tomography.

Diffuse optical tomography (DOT) is a non-invasive imaging technique utilizing multi-scattered light at visible and infrared wavelengths to detect anomalies in tissues. However, the DOT image reconstruction is based on solving the inverse problem, which requires massive calculations and time. In this article, for the first time, to the best of our knowledge, a simple, regression-based cascaded feed-forward deep learning neural network is derived to solve the inverse problem of DOT in compressed breast geometry. The predicted data is subsequently utilized to visualize the breast tissues and their anomalies. The dataset in this study is created using a Monte-Carlo algorithm, which simulates the light propagation in the compressed breast placed inside a parallel plate source-detector geometry (forward process). The simulated DL-DOT system's performance is evaluated using the Pearson correlation coefficient (R) and the Mean squared error (MSE) metrics. Although a comparatively smaller dataset (50 nos.) is used, our simulation results show that the developed feed-forward network algorithm to solve the inverse problem delivers an increment of ∼30% over the analytical solution approach, in terms of R. Furthermore, the proposed network's MSE outperforms that of the analytical solution's MSE by a large margin revealing the robustness of the network and the adaptability of the system for potential applications in medical settings.

Enhanced production of alkane hydroxylase from Penicillium chrysogenum SNP5 (MTCC13144) through feed-forward neural network and genetic algorithm

Alkane hydroxylase (AlkB), a membrane-bound enzyme has high industrial demand; however, its economical production remains challenging due to its intrinsic nature and co-factor dependency. In the current study, various critical process parameters for optimum production of AlkB have been optimized through feed forward neural network (FFNN) and genetic algorithm (GA) models using Penicillium chrysogenum SNP5 (MTCC13144). AlkB specific activity under preliminary un-optimized conditions i.e., 1% hexadecane, 7.4 pH, 11 days incubation time, 28 °C incubation temperature and 1 ml of inoculum size was 100 U/mg. ‘One variable at a time’ (OVAT) strategy was used to identify optimum physicochemical parameters and then its output data was fed to develop a model of FFNN with ‘6-12-1’ topology. Outputs of FFNN were further optimized through GA to minimize errors and intensify search level. This has provided superior predictive performances with 0.053 U/mg overall mean absolute percentage error (MAPE), 6.801 U/mg root mean square errors (RMSE), and 0.987 overall correlation coefficient (R). The AlkB specific activity improved by 3.5-fold, i.e., from 100 U/mg under preliminary un-optimized conditions to 351.32 U/mg under optimum physicochemical conditions obtained through FFNN-GA hybrid method, i.e., hexadecane (carbon source): 1.56% v/v, FeSO4: 0.63 mM, incubation temperature: 27.40 °C, pH: 7.38, incubation time: 12.35 days and inoculums size: 1.33 ml. The developed process would be a stepping stone to fulfill the high industrial demands of Alkane hydroxylase.

Machine learning algorithms for predicting direct-acting antiviral treatment failure in chronic hepatitis C: An HCV-TARGET analysis.

We aimed to develop and validate machine learning algorithms to predict direct-acting antiviral (DAA) treatment failure among patients with HCV infection. We used HCV-TARGET registry data to identify HCV-infected adults receiving all-oral DAA treatment and having virologic outcome. Potential pretreatment predictors (n=179) included sociodemographic, clinical characteristics, and virologic data. We applied multivariable logistic regression as well as elastic net, random forest, gradient boosting machine (GBM), and feedforward neural network machine learning algorithms to predict DAA treatment failure. Training (n=4894) and validation (n=1631) patient samples had similar sociodemographic and clinical characteristics (mean age, 57 years; 60% male; 66% White; 36% with cirrhosis). Of 6525 HCV-infected adults, 95.3% achieved sustained virologic response, whereas 4.7% experienced DAA treatment failure. In the validation sample, machine learning approaches performed similarly in predicting DAA treatment failure (C statistic [95% CI]: GBM, 0.69 [0.64-0.74]; random forest, 0.68 [0.63-0.73]; feedforward neural network, 0.66 [0.60-0.71]; elastic net, 0.64 [0.59-0.70]), and all four outperformed multivariable logistic regression (0.51 [0.46-0.57]). Using the Youden index to identify the balanced risk score threshold, GBM had 66.2% sensitivity and 65.1% specificity, and 12 individuals were needed to evaluate to identify 1 DAA treatment failure. Over 55% of patients with treatment failure were classified by the GBM in the top three risk decile subgroups (positive predictive value: 6%-14%). The top 10 GBM-identified predictors included albumin, liver enzymes (aspartate aminotransferase, alkaline phosphatase), total bilirubin levels, sex, HCV viral loads, sodium level, HCC, platelet levels, and tobacco use. Machine learning algorithms performed effectively for risk prediction and stratification of DAA treatment failure.

Modeling of copper removal from aqueous solutions by using carbon-based adsorbents derived from hazelnut and walnut shells by artificial neural network

The aim of this study is to model the removal of copper from an aqueous solution using Artificial Neural Network (ANN). Two different carbon-based materials (biochar) obtained by pyrolysis of hazelnut and walnut shells were used as adsorbents in a batch adsorption system. The surface area of biochars with a porous structure obtained from hazelnut and walnut shells were 124.347 m2/g and 256.931 m2/g, respectively. In adsorption experiments, initial copper concentration, adsorbent amount, temperature, pH, contact time and mixing speed were the adsorption parameters considered for the system and defined as inputs in the modeling studies. Experimental results for removing copper ions were found by changing pH 2.5-5, initial heavy metal concentration 15-45 ppm, adsorbent amount 1-3 g/L, mixing speed 200-600 rpm and temperature between 25-45oC. The % copper removal, which is tried to be maximized during the modeling phase, was selected as the model estimation parameter and defined as output to the system. ANN training was done with the Levenberg–Marquardt (LM) feed-forward algorithm and the data was categorised into 50% training, 25% validation and 25% testing. The maximum epoch value was determined as 8 iterations. Correlation coefficient (R2) values of the system were determined as 99% for education, validation and testing for the two different carbon-based adsorbents.

COMPATIVE ANALYSIS OF DEEP LEARNING MODELS FOR DDOS ATTACKS DETECTION

Recently, Distributed Denial of Service(DDOS) attacks have been on the rise and come in very many forms costing many technology firms a lot of time and money. In this study, deep learning models were compared in terms of performance, to solve the problem of detecting these attacks. The first step to mitigating DDOS attacks is by first identifying them, which serves as a toll order. This report used two deep learning models: the Deep Feed Forward (DFF) algorithm and a hybrid containing a CNN with BiLSTM (bidirectional long short-term memory). To compare these algorithms, the “DDoS Botnet Attack on IoT a71a0b42-4” dataset available on Kaggle was chosen. The dataset was undergone various evaluations to find out the performance metrics between the two algorithms. From the simulations conducted, DFF was found to have an accuracy of 87.2% with detecting the time of 0.8 seconds, while the CNN-Bi-LSTM was found to have an accuracy of 94.6% with detecting the time of 1.4 seconds.

ANN model for detection and classification of sleep and non-sleep stages

Disorders related to sleep has become one of the prime issues in human life, which may become the major reasons for health consequences. This work proposes an efficient approach to discriminate sleep stage from non-sleep stage by analysing electroencephalogram (EEG) signals from frontal lobes of brain from a single channel. Second order FIR filter is designed to separate Delta and Theta waves from the collected EEG data, empirical mode decomposition decomposes the EEG signals into small definite components which extracts distinct features like Kurtosis, median absolute deviation (MAD) and inter quartile range (IQR). An artificial neural network model is developed to classify the features of sleep and non-sleep classes which combine the advantages of feed forward and backpropagation algorithms. The experimental results obtained determine that the features extracted proved to be good discriminators of sleep and non-sleep stage with an accuracy of 92%, sensitivity 100% and specificity 98%.

Affine-Projection-Like Maximum Correntropy Criteria Algorithm for Robust Active Noise Control

Impulse noise (IN) can seriously degrade the performance of conventional active noise control (ANC) algorithms. Here, we present a novel feedforward ANC algorithm based on an information-theoretic learning framework with the data-reuse scheme of affine-projection-based algorithms. Inspired by the maximum correntropy criterion (MCC), the proposed algorithm is referred to as the modified filtered-x affine-projection-like MCC (MFxAPLMCC). Furthermore, we developed an objective function to maximize the correntropy between the system&#x0027;s desired vectors and the secondary path&#x0027;s output vectors to enhance robustness. Moreover, linear approximation reduces computational complexity, and the optimal step size is derived mathematically to accelerate convergence and increase the noise reduction ratio. The MFxAPLMCC algorithm was thoroughly evaluated in terms of stability and computational complexity; numerical simulations were used to confirm the effectiveness in terms of average noise reduction. The efficiency of our method was verified using three types of input: (1) symmetric alpha-stable (S <inline-formula><tex-math notation="LaTeX">$\alpha$</tex-math></inline-formula> S) IN, (2) a mixture of sinusoidal and IN, and (3) in-vehicle engine acceleration noise. We also verified the tracking capability of the adaptive algorithm for a case in which the primary path changes abruptly. Furthermore, the real measured acoustic path for ANC in-ear headphone development was used to validate the proposed method in real environments. The proposed algorithm significantly outperformed comparative ANC algorithms in convergence rate and noise reduction ratio. We also confirmed that the theoretical bound for stable step size coincides with the numerical results. The parameter sensitivity of the MFxAPLMCC was analyzed as well.