LMS Algorithm

This paper shows results obtained in the Automatic Speech Recognition (ASR) task for a corpus of digits speech files with a determinate noise level immerse. The experiments realized treated with several speech files that contained Gaussian noise. We used HTK (Hidden Markov Model Toolkit) software of Cambridge University in the experiments. The noise level added to the speech signals was varying from fifteen to forty dB increased by a step of 5 units. We used an adaptive filtering to reduce the level noise (it was based in the Least Measure Square -LMS-algorithm). With LMS we obtained an error rate lower than if it was not present. It was obtained because of we trained with 50% of contaminated and originals signals to the ASR. The results showed in this paper to analyze the ASR performance in a noisy environment and to demonstrate that if we have controlling the noise level and if we know the application where it is going to work, then we can obtain a better response in the ASR tasks. Is very interesting to count with these results because speech signal that we can find in a real experiment (extracted from an environment work, i.e.), could be treated with these technique and decrease the error rate obtained. Finally, we report a recognition rate of 99%, 97.5% 96%, 90.5%, 81% and 78.5% obtained from 15, 20, 25, 30, 35 and 40 noise levels, respectively when the corpus that we mentioned above was employed. Finally, we made experiments with a total of 2600 sentences (between noisy and filtered sentences) of speech signal.

A new approach for matrix inversion is introduced. The approach is based on vector representation of multiple-input multiple-output (MIMO) channel matrix, in which the channel matrix is described by a linear combination of channel vectors weighted by their respective system inputs. The MIMO system output is then fed into a bank of adaptive filters, wherein the response of a given adaptive filter is iteratively minimized to match its output to the given system input. In doing so, adaptive filters equalize the impact of respective channel vectors on the MIMO channel output, while simultaneously orthogonalizing themselves from all other channel vectors, forming the channel matrix inverse. The method demonstrates satisfactory convergence and accuracy in Monte Carlo simulations conducted with varying signal-to-noise ratios (SNRs) and matrix conditioning scenarios. The suggested approach, by virtue of its adaptable characteristics, can also be employed for time-varying linear equation systems.

A crucial performance measure in the context of the problem of deconvolution is the level of residual intersymbol interference (ISI), a metric that is classically understood and formulated in terms of an L2-norm perspective. In order to enhance the scope of ISI quantification, we propose a novel entropy-based performance measure, which is called Entropybased Intersymbol Interference (HISI). Interestingly, this metric is related to an information-theoretic relationship between source distribution and Lp-norms when the error entropy is used as a basis for optimal filtering. The new metric is analytically investigated and illustrated with some simulations.

Three Methods are discussed for controlling behavior of Arrays.The first, Schelkunoff Polynomial (Null Placement) uses information on the numbers of nulls hence the number of elements and their excitations coefficient are derived.The second is statistically based and minimizes the Mean Square Error cost function between the desired output (a Known Signal) and what the array actual outputs.To do this,it is required that the direction of arrival of the desired signal and its expected power are known.The Third one,The Least Mean Square Algorithm(Adaptive Antenna Theory) employs adaptive weighting algorithm that adapts the weights based on the received signals to improve the performance of the array.The weights,minimizes the Mean-Squared Error between a desired signal and the arrays Outputs.

This paper introduces several new least mean-square (LMS) algorithms based on error normalization procedure. Different minimization approaches and techniques were used in developing the proposed algorithms. Some of these algorithms are selected and applied to an adaptive noise canceller setup with different stationary noise power levels. Simulation results, carried out using a real speech, clearly demonstrate the superiority of the proposed algorithms over other standard LMS algorithms. Smaller values of steady-state excess mean-square error and better tracking capabilities are obtained using these new algorithms.

The main theme of this paper is to reduce noise from the noisy composite signal and reconstruct the input signals from the composite signal by designing FIR digital filter bank. In this work, three sinusoidal signals of different frequencies and amplitudes are combined to get composite signal and a low frequency noise signal is added with the composite signal to get noisy composite signal. Finally noisy composite signal is filtered by using FIR digital filter bank to reduce noise and reconstruct the input signals.

A hierarchical design methodology for wide bandwidth photoreceiver front-ends is presented in this paper. We propose a unified approach for both optoelectronic and electronic components in the front-end. This approach enables us to improve system performance by considering design constraints in both domains, which is impossible with traditional segregated design flows. Based on a hierarchical synthesis platform aimed at fast multi-domain system design automation, we optimize the photodiode structure and CMOS transimpedance amplifiers for different technology nodes with accurate performance prediction.

While the LMS algorithm and its normalized version (NLMS), have been thoroughly used and studied. Connections between the Kalman filter and the RLS algorithm have bean established however, the connection between the Kalman filter and the LMS algorithm has not received much attention. By linking these two algorithms, a new normalized Kalman based LMS (KLMS) algorithm can be derived that has some advantages to the classical one. Their stability is guaranteed since they are a special case of the Kalman filter. More, they suggests a new way to control the step size, that results in good convergence properties for a large range of input signal powers, that occur in many applications. They prevent high measurement noise sensitivity that may occur in the NLMS algorithm for low order filters, like the ones used in OFDM equalization systems. In these paper, different algorithms based on the correlation form, information form and simplified versions of these are presented. The simplified form maintain the good convergence properties of the KLMS with slightly lower computational complexity. 1

While the LMS algorithm and its normalized version (NLMS), have been thoroughly used and studied, and connections between the Kalman filter and the RLS algorithm have bean established, the connection between the Kalman filter and the LMS algorithm has not received much attention. By linking these two algorithms, a new normalized LMS algorithm can be derived that has some advantages to the classical one. Firstly, its stability is guaranteed since it is a special case of the Kalman filter. Secondly, it suggests a new way to control the step size that results in optimum convergence for a large range of input signal powers, that occur in many applications. Finally, it prevents measurement noise amplification that occurs in the NLMS algorithm for low order filters, like the ones used in OFDM equalization systems.

In this letter, we propose a novel least-mean-square (LMS) algorithm for filtering speech sounds in the adaptive noise cancellation (ANC) problem. It is based on the minimization of the squared Euclidean norm of the difference weight vector under a stability constraint defined over the a posteriori estimation error. To this purpose, the Lagrangian methodology has been used in order to propose a nonlinear adaptation rule defined in terms of the product of differential inputs and errors which means a generalization of the normalized (N)LMS algorithm. The proposed method yields better tracking ability in this context as shown in the experiments which are carried out on the AURORA 2 and 3 speech databases. They provide an extensive performance evaluation along with an exhaustive comparison to standard LMS algorithms with almost the same computational load, including the NLMS and other recently reported LMS algorithms such as the modified (M)-NLMS, the error nonlinearity (EN)-LMS, or the normalized data nonlinearity (NDN)-LMS adaptation.

Gradient descent method is one of the popular methods to train feedforward neural networks. Batch and incremental modes are the two most common methods to practically implement the gradient-based training for such networks. Furthermore, since generalization is an important property and quality criterion of a trained network, pruning algorithms with the addition of regularization terms have been widely used as an efficient way to achieve good generalization. In this paper, we review the convergence property and other performance aspects of recently researched training approaches based on different penalization terms. In addition, we show the smoothing approximation tricks when the penalty term is non-differentiable at origin.

Gradient descent method is one of the popular methods to train feedforward neural networks. Batch and incremental modes are the two most common methods to practically implement the gradient-based training for such networks. Furthermore, since generalization is an important property and quality criterion of a trained network, pruning algorithms with the addition of regularization terms have been widely used as an efficient way to achieve good generalization. In this paper, we review the convergence property and other performance aspects of recently researched training approaches based on different penalization terms. In addition, we show the smoothing approximation tricks when the penalty term is non-differentiable at origin.

Non-negativity constraints arise naturally in many applications such as medical and astronomical imaging. These constraints have been recently exploited in the optical communication field where a projected parallel interference cancelation detector has been developed within the context of cancelling multi-access interference in upstream on-off-keying optical code division multiple access passive optical networks; the latter possess the property that the transmitted data symbols are nonnegative. The authors build upon these results and propose a new parallel interference cancellation structure with better features and improved performance. This detector is characterised by a multiplicative update equation and it exhibits a number of unique and desirable features such as stability and smooth convergence behaviour even for highly loaded systems. Up to our knowledge, this structure has not been proposed before either in optical or in wireless communication field. Simulation results are promising and further improvements are possible.

In this paper we study three recursive identification algorithms with respect to their tracking capability and disturbance sensitivity. The algorithms are applied to the identification of time-varying systems with finite impulse response. The performance of the algorithms are measured in terms of the mean square error of the estimated transfer function. Using asymptotic methods we derive explicit expressions for the mean square error, and these expressions are used to evaluate the performance of the algorithms when applied to systems with different behavior.

This paper presented a study of three algorithms, the<br> equalization algorithm to equalize the transmission channel with ZF<br> and MMSE criteria, application of channel Bran A, and adaptive<br> filtering algorithms LMS and RLS to estimate the parameters of the<br> equalizer filter, i.e. move to the channel estimation and therefore<br> reflect the temporal variations of the channel, and reduce the error in<br> the transmitted signal. So far the performance of the algorithm<br> equalizer with ZF and MMSE criteria both in the case without noise,<br> a comparison of performance of the LMS and RLS algorithm.

A major challenge in acoustic feedback cancellation is the strong correlation between the excitation signal and the error signal, caused by the closed electro-acoustic loop. Due to this correlation, the convergence rate of adaptive algorithms, such as the NLMS, is limited. It was shown in a recent publication that the convergence can be improved to a large extent by using a reverb-based stepsize control. This stepsize control aims at increasing the stepsize during reverberant signal periods and decreasing the stepsize during local speech activity. Caused by reverberation of the acoustic system, there is still energy in the system immediately after local speech periods. This reverberation can be exploited to adapt the filter, since the signals are not correlated here. In this paper, the reverbbased stepsize control is further improved. Therefore, the gain in the forward path of the closed-loop system is controlled with the system distance. It is shown that in this case, the reverb-ba...

Time-varying system behavior learning is crucial in the current landscape of machine learning, where statistical behavior often lacks precision. This approach effectively models many real-world scenarios characterized by their time-varying non-stationary nature. The focus has shifted towards adaptive systems for these applications, as they demonstrate superior performance compared to traditional fixed-weight systems that have been in use for a longer period. The introduction of multilayer adaptation has further enhanced the effectiveness of signal processing in this context. A single layer of adaptation utilizing a recursive or steepest-gradient filter is insufficient to ensure both rapid convergence and optimal error fidelity during the learning process, particularly in unknown time-varying environments. Therefore, a multilayer adaptation approach that combines least mean squares (LMS) and recursive least squares (RLS) algorithms has been proposed, incorporating a mixing parameter [5], [6], [7] that plays a pivotal role in the learning process. This multilayer adaptation has been shown to significantly improve learning capabilities, achieving approximately 4 to 5 dB better noise fidelity compared to existing single-layer adaptation methods.

We would like to express our gratitude towards all the people who have contributed their precious time and effort to help me. Without whom it would not have been possible for us to understand and complete the project. We would like to thank Prof Sanjeeb Mohanty, Department of Electrical Engineering, and our Project Supervisor for his guidance, support, motivation and encouragement for this project work. His readiness for consultation at all times, his educative comments, his concern and assistance even with practical things have been invaluable. We are grateful to Prof. A. K. Panda, Head of the Department of Electrical Engineering for providing necessary facilities in the department.

We proposed a blind dereverberation method based on spectral subtraction by Multi-Channel Least Mean Square (MCLMS) algorithm for distant-talking speech recognition in our previous study . In this paper, we discuss the problems of the proposed method and present some solutions. In a distant-talking environment, the length of channel impulse response is longer than the short-term spectral analysis window. By treating the late reverberation as additive noise, a noise reduction technique based on spectral subtraction was proposed to estimate power spectrum of the clean speech using power spectra of the distorted speech and the unknown impulse responses. To estimate the power spectra of the impulse responses, a Variable Step-Size Unconstrained MCLMS (VSS-UMCLMS) algorithm for identifying the impulse responses in a time domain was extended to a frequency domain. To reduce the effect of the estimation error of channel impulse response, we normalize the early reverberation by CMN instead of the spectral subtraction used by the estimated impulse response in this paper. Furthermore, our proposed method is combined with a conventional delay-andsum beamforming. We conducted the experiments on distorted speech signal simulated by convolving multi-channel impulse responses with clean speech. The modified proposed method achieved a relative error reduction rate of 22.7% from conventional CMN and 12.0% from the original proposed method, respectively. By combining the modified proposed method with the beamforming, a furthermore improvement (relative error reduction rate of 23.3%) was achieved.

We propose a blind dereverberation method based on spectral subtraction using a multi-channel least mean squares (MCLMS) algorithm for distant-talking speech recognition. In a distant-talking environment, the channel impulse response is longer than the short-term spectral analysis window. By treating the late reverberation as additive noise, a noise reduction technique based on spectral subtraction was proposed to estimate the power spectrum of the clean speech using power spectra of the distorted speech and the unknown impulse responses. To estimate the power spectra of the impulse responses, a variable step-size unconstrained MCLMS (VSS-UMCLMS) algorithm for identifying the impulse responses in a time domain is extended to a frequency domain. To reduce the effect of the estimation error of the channel impulse response, we normalize the early reverberation by cepstral mean normalization (CMN) instead of spectral subtraction using the estimated impulse response. Furthermore, our proposed method is combined with conventional delay-andsum beamforming. We conducted recognition experiments on a distorted speech signal simulated by convolving multi-channel impulse responses with clean speech. The proposed method achieved a relative error reduction rate of 22.4% in relation to conventional CMN. By combining the proposed method with beamforming, a relative error reduction rate of 24.5% in relation to the conventional CMN with beamforming was achieved using only an isolated word (with duration of about 0.6 s) to estimate the spectrum of the impulse response.

Adaptive ¯lters have wide range of applications in areas such as echo or interference cancellation, prediction and system identi¯cation. Due to high computational complexity of adaptive ¯lters, their hardware implementation is not an easy task. However, it becomes essential in many cases where real-time execution is needed. This paper presents the design and hardware implementation of a variable step size 40 order adaptive ¯lter for de-noising acoustic signals. To ensure an area e±cient implementation, a novel structure is being proposed. The proposed structure eliminates the requirement of extra registers for storage of delayed inputs thereby reducing the silicon area. The structure is compared with direct-form and transposed-form structures by adapting the ¯lter coe±cients using four di®erent variants of the least means square (LMS) algorithm. Subsequently, the ¯lters are implemented on three di®erent ¯eld programmable gate arrays (FPGAs) viz. Spartan 6, Virtex 6 and Virtex 7 to ¯nd out the best device family that can be used to implement an Adaptive noise canceller (ANC) by comparing speed, power and area utilization. The synthesis results clearly reveal that ANC designed using the proposed structure has resulted in a reduction in silicon area without incurring any sig-ni¯cant overhead in terms of power or delay.

It is possible to obtain the porosity and permeability of filter cakes formed during cake filtration by fitting power law type relationships to experimental data. Specifically, local filtration data of the form of pressure and concentration profiles can be used to determine the constitutive relationships. However, when applied to data on the filtration of suspensions encountered in the pulp and paper industry, methods of minimization such as the Levenberg-Marquardt (LM) method fail to produce unique sets of model parameters representing the constitutive relationships. Recourse to more sophisticated methods such as simulated annealing (SA) has to be made in order to obtain suitable global minima for describing wide ranging experimental data. In this paper, we chose experimental filtration data of pulp mill lime muds (essentially CaCO 3 particulate suspensions) reported by 'Sep. Technol. 5 (1995) 165' to study the best means of obtaining robust parameter sets for use in the permeability and compressibility relationships. The efficacy of a heuristic algorithm based on SA over a straightforward LM algorithm as applied to the filtration of lime muds and papermaking pulps is demonstrated. A second example of constant pressure filtration of papermaking pulps reported by 'TAPPI 35 (1952) 439' was also chosen. The parameters in the permeability relationship for pulp mats consist of the specific surface area (hydrodynamic) and the specific volume of the papermaking fibers. By using the SA method, it is shown that these two parameters as well as the compressibility parameters could be determined directly.

We propose in this paper a new adaptive algorithm, designed to track system impulse responses, characterized by stochastic Markovian time variations. The proposed nonstationary least mean square (NSLMS) algorithm is designed so that it explicitly takes into account the structure of the nonstationarity. Hence, unlike the classical LMS algorithm, the NSLMS algorithm is not blind with respect to the time variations of the system impulse response to identify. The proposed algorithm structure is based on a coupling between the estimation of the Markovian parameter that characterizes the nonstationarity and the estimation of the adaptive filter that identifies the system impulse response. The adaptive identification of the Markovian parameter is performed by an LMS algorithm, based on the minimization of the mean square of the system identification error. A theoretical analysis of the transient and the steady-state behaviors of the NSLMS adaptive filter is carried out. In particular, an analytical expression of the step size that guarantees the stability of the latter is established. The theoretical misadjustment that measures the tracking ability of the NSLMS algorithm is computed for an i.i.d. input. We prove that in the steady-state, the NSLMS algorithm exhibits better performance than the classical LMS algorithm, and goes beyond the limitation of the LMS algorithm to track severe filter time variations. The experimental results reported here are in perfect agreement with the theory. They display the good properties of the NSLMS algorithm and demonstrate its ability to yield good performances in a hard Markovian time varying environment.

In this paper a modification on Levenberg-Marquardt algorithm for MLP neural network learning is proposed. The proposed algorithm has good convergence. This method reduces the amount of oscillation in learning procedure. An example is given to show usefulness of this method. Finally a simulation verifies the results of proposed method.

In this paper a modification on Levenberg-Marquardt algorithm for MLP neural network learning is proposed. The proposed algorithm has good convergence. This method reduces the amount of oscillation in learning procedure. An example is given to show usefulness of this method. Finally a simulation verifies the results of proposed method.

This paper describes the OCDMA technology (Optical Code Division Multiple Access) focussing on the coherent systems OCDMA. The modelling was done using VPI v7.0 and the results of the simulations are presented. The utility of OCDMA as a control access technology it's demonstrated by the simulations which describe the most important aspect to be considered to evaluate the performance of coherent OCDMA Systems.

Asymmetric digital subscriber lines (ADSLs) employ discrete multitone modulation (DMT) as transmission format, where subcarriers are assigned to the up-and/or downstream transmission direction. To separate up-and downstream signals, the ADSL standard allows the use of echo cancellation resulting in improved bit rates, reach, and/or noise margins. In DMT-based modems, typically, the mixed time/frequency (MTF) domain echo canceling scheme, as proposed by Ho et al., is implemented. This technique estimates the echo filter in the frequency domain using the least mean square (LMS) algorithm with the transmitted echo symbols as update directions. Since not every tone of the transmitted echo signal will carry data, i.e., will be excited, the MTF adaptation process does not lead to a good estimate for the echo channel, unless extra power on unused echo tones is transmitted. However, transmitting extra power on such tones is often undesired. In this paper, we present an alternative echo canceling scheme referred to as the circulant decomposition canceler (CDC), which works without extra power requirements and with comparable complexity as the method of Ho et al. Similar to MTF echo canceling, the CDC scheme can easily be incorporated into a multirate environment with different transmit and receive rates and can also cheaply be combined with per-tone equalization and double talk cancellation to allow fast tracking and/or convergence in the presence of a far-end signal.

In this paper an analog adaptive filter circuit is presented, their coefficients are adapted with analog LMS algorithm using CMOS HP 0.5um technology. The layout simulation result shows fast convergence speed and low power consumption. The performance in configuration as an identifier shows that this circuit can be used in portables VLSI communications systems.

In this paper an analog adaptive filter circuit is presented, their coefficients are adapted with analog LMS algorithm using CMOS HP 0.5um technology. The layout simulation result shows fast convergence speed and low power consumption. The performance in configuration as an identifier shows that this circuit can be used in portables VLSI communications systems.

The conventional algorithms in the echo canceling system have drawback when they are faced with double-talk condition in noisy environment. Since the double-talk and noise signal are exist, then the error signal is contaminated to estimate the gradient correctly. In this paper, we define a new class of adaptive algorithm for tap adaptations, based on the correlation function processing. The computer simulation results show that the Correlation LMS (CLMS) and the Extended CLMS (ECLMS) algorithms have better performance than conventional LMS algorithm. In order to implement the ECLMS algorithm, the Frequency domain Extended CLMS (FECLMS) algorithm is proposed to reduce the computational complexity. However the convergence speed is not sufficient. In order to improve the convergence speed, the Wavelet domain Extended CLMS (WE-CLMS) algorithm is proposed. The computer simulation results support the theoretical findings and verify the robustness of the proposed WECLMS algorithm in the double-talk situation.

In this paper, the idea of multi-split adaptive filtering is applied on DFE. The feed-forward and feedback sections in the DFE are divided into parallel sub-filters by imposing separately the symmetry and antisymmetry conditions on the impulse responses of filters by using appropriate and distinct sets of linear constraints. The purpose of this is to obtain a new structure that gives better performance than the conventional structure. The performance of multi-split DFE is tested with different algorithms (LMS, RLS, DCT-LMS) and compared to that of multi-split linear transversal Equalizer. convergence and complexity of the equalizer are two conflicting parameters. In order to improve the convergence rate with a small increase in complexity, the split adaptive filter is a good choice [4][5][6]. Split processing technique guarantees stability and give quicker convergence at the expense of a moderate increase in computation. In adaptive equalization, the convergence behavior is governed by the eigenvalue spread; a lower spread increases the convergence rate. It has been shown that whitening the equalizer input using some transformations such as Karhumen-Loeve transform and discrete cosine transform can significantly reduce the eigenvalue spread [4].

In this paper, we present a novel modification to the standard particle swarm optimization (PSO) technique and illustrate the superiority of the proposed modified technique over other PSO-based techniques, with an application to the important area of adaptive channel equalization. Different published versions of the original PSO algorithms are first reviewed and the new proposed technique discussed in the context of the design of adaptive channel equalizers. An exhaustive simulation-based sensitivity analysis of the proposed PSO algorithm, with respect to its underpinning parameters, is carried out here so as to select the "best" (or near optimal) values of these parameters. The performance of various PSO algorithms, including our proposed algorithm, is compared in the context of adaptive channel equalization to that of the LMS algorithm through extensive simulations. This detailed comparison revealed the superior performance of our proposed PSO-based adaptive channel equalizer over both its LMS-based counterpart and other adaptive equalizers based on the published PSO algorithms. This superior performance was exhibited on both linear and nonlinear channels.

Sound radiation of planar radiators such as beams and plates is known to be directly related to the velocity distribution over the structural surface at low frequencies. For example, nonvolumetric modes correspond to poor sound radiators for small k0a. In this work, to achieve significant sound attenuation in the low-frequency range, the SISO eigenassignment technique is used to modify the eigenproperties of a planar structure using structural actuators and sensors so that all the modes of the controlled structure are nonvolumetric. The main advantage of such an approach is that the design is independent of the disturbance characteristics (i.e., type, position, and frequency content) and does not require sensors in the radiation field. The design procedure for the control system in the modal domain is presented. The formulation is applied to a simply supported beam with SISO feedforward control. Radiation efficiency, far-field sound radiated power and mean-square surface velocity ar...

C-slow retiming is a process of automatically increasing the throughput of a design by enabling fine grained pipelining of problems with feedback loops. This transformation is especially appropriate when applied to FPGA designs because of the large number of available registers. To demonstrate and evaluate the benefits of C-slow retiming, we constructed an automatic tool which modifies designs targeting the Xilinx Virtex family of FPGAs. Applying our tool to three benchmarks: AES encryption, Smith/Waterman sequence matching, and the LEON 1 synthesized microprocessor core, we were able to substantially increase the total throughput. For some parameters, throughput is effectively doubled.

Recently, we have proposed an adaptive channel estimation (CE) scheme using one-tap recursive least square (RLS) algorithm (adaptive RLS-CE), where the forgetting factor is adapted to the changing channel condition by the least mean square (LMS) algorithm, for direct sequence-code division multiple access (DS-CDMA) with frequency-domain equalization (FDE). However, the tracking ability for adaptive RLS-CE is limited since the channel estimate obtained from the previously received block is used. In this paper, we introduce the polynomial prediction to improve the tracking ability. We evaluate the bit error rate (BER) performance of DS-CDMA using polynomial prediction RLS-CE in a frequency-selective fast Rayleigh fading channel by computer simulation.

Face milling is performed on aluminum alloy A96061-T6 at diverse cutting parameters proposed by the design of experiments. Surface roughness is predicted by examining the effects of cutting parameters (CP), vibrations (Vib), and sound characteristics (SC). Sound characteristics based on surface roughness estimation determine the rarity of the work. In this study, a unique ANN-TLBO hybrid model (Artificial Neural Networks: Teaching Learning Based Algorithm) is created to predict the surface roughness from CP, Vib, and SC. To ascertain their correctness and efficacy in evaluating surface roughness, the performance of these models is evaluated. First off, the CP hybrid model demonstrated an amazing accuracy of 95.1%, demonstrating its capacity to offer trustworthy forecasts of surface roughness values. The Vib hybrid model, in addition, demonstrated a respectable accuracy of 85.4%. Although it was not as accurate as the CP model, it nevertheless showed promise in forecasting surface roughness. The SC-based hybrid model outperformed the other two models in terms of accuracy with a remarkable accuracy of 96.2%, making it the most trustworthy and efficient technique for assessing surface roughness in this investigation. An analysis of error percentages revealed the exceptional performance of SC-based Model-3, exhibiting an average error percentage of 3.77%. This outperformed Vib Model-2 (14.52%) and CP-based Model-1 (4.75%). The SC model is the best option, and given its outstanding accuracy, it may end up becoming the go-to technique for industrial applications needing accurate surface roughness measurement. The SC model's exceptional performance highlights the importance of optimization strategies in improving the prediction capacities of ANN-based models, leading to significant advancements in the field of surface roughness assessment and related fields. An IoT platform is developed to link the model's output with other systems. The system created eliminates the need for manual, physical surface roughness measurement and allows for the display of surface roughness data on the cloud and other platforms.

Over the last decade, kernel methods for nonlinear processing have successfully been used in the machine learning community. However, so far, the emphasis has been on batch techniques. It is only recently, that online adaptive techniques have been considered in the context of signal processing tasks. To the best of our knowledge, no kernel-based strategy has been developed, so far, that is able to deal with complex valued signals. In this paper, we take advantage of a technique called complexification of real RKHSs to attack this problem. In order to derive gradients and subgradients of operators that need to be defined on the associated complex RKHSs, we employ the powerful tool of Wirtinger's Calculus, which has recently attracted much attention in the signal processing community. Writinger's calculus simplifies computations and offers an elegant tool for treating complex signals. To this end, in this paper, the notion of Writinger's calculus is extended, for the first time, to include complex RKHSs and use it to derive the Complex Kernel Least-Mean-Square (CKLMS) algorithm. Experiments verify that the CKLMS can be used to derive nonlinear stable algorithms, which offer significant performance improvements over the traditional complex LMS or Widely Linear complex LMS (WL-LMS) algorithms, when dealing with nonlinearities.

This paper proposes a new Least Mean Square (LMS) based algorithm aimed for acoustic echo cancellation. The algorithm is an elaboration of an existing algorithm developed for the Konftel 200 conference phone. The purpose of the algorithm is to reduce the effects of quantization with only a small increase in computational load as compared to the LMS algorithm. The algorithm makes use of the specific transfer characteristic of certain systems, e.g. hands-free phones, in combination with finite wordlength precision arithmetic. It is shown, both analytically and with simulations that the new algorithm outperforms the LMS algorithm when implemented in a two's-complement arithmetic system identification scheme.

This paper presents a brushless dc motor drive for heating, ventilating and air conditioning fans, which is utilized as the load of a photovoltaic system with a maximum power point tracking (MPPT) controller. The MPPT controller is based on a genetic assisted, multi-layer perceptron neural network (GA-MLP-NN) structure and includes a DC-DC boost converter. Genetic assistance in the neural network is used to optimize the size of the hidden layer. Also, for training the network, a genetic assisted, Levenberg-Marquardt (GA-LM) algorithm is utilized. The off line GA-MLP-NN, trained by this hybrid algorithm, is utilized for online estimation of the voltage and current values in the maximum power point. A brushless dc (BLDC) motor drive system that incorporates a motor controller with proportional integral (PI) speed control loop is successfully implemented to operate the fans. The digital signal processor (DSP) based unit provides rapid achievement of the MPPT and current control of the BLDC motor drive. The performance results of the system are given, and experimental results are presented for a laboratory prototype of 120 W.

In this paper, we are to develop a second-order LMS-type Volterra filter to reduce distortions of data transmission over analog telephone channels due the channel impulse response and inter-symbol interference (ISI). A novel approach for updating the linear and quadratic coefficients vectors of a second-order Volterra filter is presented. The innovative features of this algorithm provide distinct convergence treatment of the linear and quadratic Volterra filter kernels, and adaptive formulation for the convergence factors, hence providing a unique platform for both linear and nonlinear channel adaptive handling. Simulations are carried in an equalization set-up to compare the performance of this algorithm with conventional and variable step LMS algorithms. The obtained results show that this algorithm brings substantial increase in the convergence speed while keeping, to some extent, the simplicity of the conventional LMS algorithm. Considering the performance of the formulated algorithm, it could be a useful tool for adaptive equalization applications.

Abstract—In recent years, Computer Aided Design (CAD) based on Artificial Neural Networks (ANNs) have been introduced for microwave modeling, simulation and optimization. In this paper, the characteristic parameters of edge coupled and conductor-backed edge coupled Coplanar Waveguides have been determined with the use of ANN model. Eight learning algorithms, Levenberg-Marquart (LM),

In this paper a new adaptive complex digital filter structure is proposed. First, a very low sensitivity second-order complex bandpass (BP) filter section with independent tuning of the central frequency and the bandwidth (BW) is developed (the low sensitivity of the narrow-band realization is ensuring a higher BW-tuning accuracy or better precision in a severe coefficient quantization). Then, a BP/Bandstop (BS) adaptive filter structure is formed around this section, using LMS algorithm to adapt the central frequency. The developed filter circuit is providing a low computational complexity and a very fast convergence, and is convenient for cancellation/enhancement of complex sinusoids or narrowband signals. All theoretical results in the work are verified experimentally.

Abstract: The method of steepest-descent is re-visited in continuous time. It is shown that the continuous time version is a vector differential equation the solution of which is found by integration. Since numerical integration has many forms, we show an alternative to the conventional solution by using a Trapezoidal integration solution. This in turn gives a slightly modified least-mean squares (LMS) algorithm.

Two new discrete-time algorithms are presented for tracking variance and reciprocal variance. The closed loop nature of the solutions to these problems makes this approach highly accurate and can be used recursively in real time. Since the Least-Mean Squares (LMS) method of parameter estimation requires an estimate of variance to compute the step size, this technique is well suited to applications such as speech processing and adaptive filtering.