Design of twodimensional digital filters using neural networks

A new approach for the design of two-dimensional （2-D） linear phase FIR digital filters based on a new neural networks algorithm （NNA） is provided. A compact expression for the transfer function of a 2-D linear phase FIR filter is derived based on its frequency response characteristic, and the NNA, based on minimizing the square-error in the frequency-domain, is established according to the compact expression. To illustrate the stability of the NNA, the convergence theorem is presented and proved. Design examples are also given, and the results show that the ripple is considerably small in passband and stopband, and the NNA-based method is of powerful stability and requires quite little amount of computations.

ADAPTIVE RECURRENT NEURAL NETWORKS TRACKING-FILTER FOR MANEUVERING TARGET

It is a challenge to track the maneuvering targets with noise disturbance and unknown dynamics. In this paper, an adaptive recurrent neural network tracking filter (ARNNF) for use in maneuvering target tracking was provided. The scheme is based on recurrent neural networks of which the recurrence provides a potentially unlimited memory depth adjusted by the network adaptively ( i.e. , it finds the best duration to represent the input signals past), and thus can actually capture the dynamics of the system that produced a temporal signal. On the other hand, recurrent neural network can approximate arbitrary nonlinear functions in L 2 space. The theoretical analysis indicates that the ARNNF can track the maneuvering targets with optimal filtering performance. Comparisons with IMM and AIMM algorithm show that ARNNF has better performance, and furthermore the ARNNF does not rely on the assumption with the known maneuvering target models, measurement noise and system noise.

Fuzzy neural network image filter based on GA

A new nonlinear image filter using fuzzy neural network based on genetic algorithm is proposed. The learning of network parameters is performed by genetic algorithm with the efficient binary encoding scheme. In the following, fuzzy reasoning embedded in the network aims at restoring noisy pixels without degrading the quality of fine details. It is shown by experiments that the filter is very effective in removing impulse noise and significantly outperforms conventional filters.

Filtering and Estimation of Vehicular Dead Reckoning System Based on Hopfield Neural Network

The algorithm of Hopfield neural network filtering and estimation is studied. The model of vehicular dead reckoning system fitting for the algorithm is constructed, and the design scheme of system filtering and estimation based on Hopfield network is proposed. Compared with Kalman filter, the algorithm does not require very precise system model and the prior knowledge of noise statistics and does not diverge easily. The simulation results show that the vehicular dead reckoning system based on Hopfield network filtering and estimation has the good position precision, and needn't require the inertial sensors with high precision. Therefore, the algorithm has the good practicability.

Simulation of Cellular Neural Networks by Wave Digital Filter Principles

Based on wave digital filter(WDF) principles, this paper presents a digital model of cellular neural networks(CNNs). The model can precisely simulate the dynamic behavior of CNNs.

Kalman Filters versus Neural Networks in Battery State-of-Charge Estimation: A Comparative Study

Battery management systems (BMS) must estimate the state-of-charge (SOC) of the battery accurately to prolong its lifetime and ensure a reliable operation. Since batteries have a wide range of applications, the SOC estimation requirements and methods vary from an application to another. This paper compares two SOC estimation methods, namely extended Kalman filters (EKF) and artificial neural networks (ANN). EKF is a nonlinear optimal estimator that is used to estimate the inner state of a nonlinear dynamic system using a state-space model. On the other hand, ANN is a mathematical model that consists of interconnected artificial neurons inspired by biological neural networks and is used to predict the output of a dynamic system based on some historical data of that system. A pulse-discharge test was performed on a commercial lithium-ion (Li-ion) battery cell in order to collect data to evaluate those methods. Results are presented and compared.

Optimal design study of high order FIR digital filters based on neural network algorithm

An optimal design approach of high order FIR digital filter is developed based on the algorithm of neural networks with cosine basis function . The main idea is to minimize the sum of the square errors between the amplitude response of the desired FIR filter and that of the designed by training the weights of neural networks, then obtains the impulse response of FIR digital filter . The convergence theorem of the neural networks algorithm is presented and proved, and the optimal design method is introduced by designing four kinds of FIR digital filters , i.e., low-pass, high-pass, bandpass , and band-stop FIR digital filter. The results of the amplitude responses show that attenuation in stop-bands is more than 60 dB with no ripple and pulse existing in pass-bands, and cutoff frequency of passband and stop-band is easily controlled precisely .The presented optimal design approach of high order FIR digital filter is significantly effective.

Predicting of the Fibrous Filters Efficiency for the Removal Particles from Gas Stream by Artificial Neural Network

In this paper, artificial neural networks are used for predicting single fiber efficiency in the process of removing smaller particles from gas stream by fiber filters. For this, numerical simulations are obtained of a classic model of literature for fiber efficiency, which is numerically solved along with the convection diffusion equation in polar coordinates for particle concentration, with associated initial and boundary conditions. A sufficient number of examples from two numerical simulations are employed to construct a database, from which parameters of a novel neural model are adjusted. This model is constructed based on the back propagation algorithm in order to map two features, namely Peclet number and packing density, which are extracted from the numerical simulations into the corresponding single fiber efficiency. The results indicate that the developed neural model can be trained in a reasonable computational time and is capable of estimating single fiber efficiency from examples of the test set with a maximum error of 1.7%.

Rudder Roll Damping Autopilot Using Dual Extended Kalman Filter–Trained Neural Networks for Ships in Waves

The roll motions of ships advancing in heavy seas have severe impacts on the safety of crews,vessels,and cargoes;thus,it must be damped.This study presents the design of a rudder roll damping autopilot by utilizing the dual extended Kalman filter(DEKF)trained radial basis function neural networks(RBFNN)for the surface vessels.The autopilot system constitutes the roll reduction controller and the yaw motion controller implemented in parallel.After analyzing the advantages of the DEKF-trained RBFNN control method theoretically,the ship’s nonlinear model with environmental disturbances was employed to verify the performance of the proposed stabilization system.Different sailing scenarios were conducted to investigate the motion responses of the ship in waves.The results demonstrate that the DEKF RBFNN based control system is efficient and practical in reducing roll motions and following the path for the ship sailing in waves only through rudder actions.

Neural network-based H∞ filtering for nonlinear systems with time-delays

A novel H∞ design methodology for a neural network-based nonlinear filtering scheme is addressed. Firstly, neural networks are employed to approximate the nonlinearities. Next, the nonlinear dynamic system is represented by the mode-dependent linear difference inclusion （LDI）. Finally, based on the LDI model, a neural network-based nonlinear filter （NNBNF） is developed to minimize the upper bound of H∞ gain index of the estimation error under some linear matrix inequality （LMI） constraints. Compared with the existing nonlinear filters, NNBNF is time-invariant and numerically tractable. The validity and applicability of the proposed approach are successfully demonstrated in an illustrative example.

Neural Network-based GPS/INS Integrated System for Spacecraft Attitude Determination

Global Positioning System （GPS） /Inertial Navigation System （INS） integrated system is continuously gaining research interests in many positioning and navigation fields. Kalman filtering-based integrated algorithm has some drawbacks on stability, computation load, robustness, and system observability performances. Based on neural network technology, a new GPS/INS integration filtering algorithm is studied for an integration scheme of the attitude determination GPS/INS integrated navigation system. Through some theoretic analysis, this algorithm not only has good estimation performance, but also has better robustness to the system model and noise than the traditional Kalman algorithm. To assess the performance of the proposed integrated model more deeply, some simulation is done to compare with the traditional Kalman filter model. The results indicate that the proposed model provides a significant improvement in some performance, such as accuracy, stability, robustness, and so on.

Recurrent neural network for vehicle dead-reckoning

For vehicle integrated navigation systems, real-time estimating states of the dead reckoning （DR） unit is much more difficult than that of the other measuring sensors under indefinite noises and nonlinear characteristics. Compared with the well known, extended Kalman filter （EKF）, a recurrent neural network is proposed for the solution, which not only improves the location precision and the adaptive ability of resisting disturbances, but also avoids calculating the analytic derivation and Jacobian matrices of the nonlinear system model. To test the performances of the recurrent neural network, these two methods are used to estimate the state of the vehicle＇s DR navigation system. Simulation results show that the recurrent neural network is superior to the EKF and is a more ideal filtering method for vehicle DR navigation.

Novel Sequential Neural Network Learning Algorithm for Function Approximation

A novel sequential neural network learning algorithm for function approximation is presented. The multi-step-ahead output predictor of the stochastic time series is introduced to the growing and pruning network for constructing network structure. And the network parameters are adjusted by the proportional differential filter （PDF） rather than EKF when the network growing criteria are not met. Experimental results show that the proposed algorithm can obtain a more compact network along with a smaller error in mean square sense than other typical sequential learning algorithms.

Rejection of Narow─Band Interference in DS Spread Spectrum Systems Using Recurent Neural Networks

Adaptive whitening filters have been proven to be powerful to eliminate narrow band interference in spread spectrum (SS) communication systems. However digital implementation of such kind adaptive filters is difficult for applications with chip rate as

RADAR TARGET IDENTIFICATION BY ADAPTIVE DISCRIMINATION WAVEFORM SYNTHESIS AND NEAREST NEIGHBOR NEURAL NETWORK

In this paper,a new radar target identification scheme is presented based on adaptivediscrimination waveform synthesis and a nearest neighbor neural network.It can directly use theimpulse response of the target to synthesize discrimination waveform,so the poles extractionprocedure is not required.Particularly,it can successfully operate on the case that the poles ofthe target are weakly dependent on the aspect angle.

Neural Network Based Order Statistic Processing Engines

Order statistic filters are used often in the applications of science and engineering problems. This paper investigates the design and training of a feed-forward neural network to approximate minimum, median and maximum operations. The design of order statistic neural network filtering (OSNNF) is further refined by converting the input vectors with elements of real numbers to a set of inputs consisting of ones and zeros, and the neural network is trained to yield a rank vector which can be used to obtain the exact ranked values of the input vector. As a case study, the OSNNF is used to improve the visibility of target echoes masked by clutter in ultrasonic nondestructive testing applications.

High-resolution Image Reconstruction by Neural Network and Its Application in Infrared Imaging

As digital image techniques have been widely used, the requirements for high-resolution images become increasingly stringent. Traditional single-frame interpolation techniques cannot add new high frequency information to the expanded images, and cannot improve resolution in deed. Multiframe-based techniques are effective ways for high-resolution image reconstruction, but their computation complexities and the difficulties in achieving image sequences limit their applications. An original method using an artificial neural network is proposed in this paper. Using the inherent merits in neural network, we can establish the mapping between high frequency components in low-resolution images and high-resolution images. Example applications and their results demonstrated the images reconstructed by our method are aesthetically and quantitatively (using the criteria of MSE and MAE) superior to the images acquired by common methods. Even for infrared images this method can give satisfactory results with high definition. In addition, a single-layer linear neural network is used in this paper, the computational complexity is very low, and this method can be realized in real time.

基于BP-DCKF-LSTM的锂离子电池SOC估计

电池荷电状态(SOC)的准确估计是电池管理系统(BMS)的核心功能之一。为了提高锂电池SOC估算精度,提出了一种将反向传播神经网络(BP)、双容积卡尔曼滤波(DCKF)和长短期记忆神经网络(LSTM)相结合的SOC估计方法。针对多温度条件下传统多项式拟合法在拟合开路电压(OCV)与SOC时效果较差的问题,提出了一种基于BP神经网络的拟合方法,通过验证表明该方法能有效提高拟合精度。针对单独使用模型法或数据驱动法估计SOC各自存在的优缺点,提出了一种将DCKF与LSTM相结合的估计方法,在提高估计精度的同时,可以减少参数调节时间和训练成本。实验验证表明,BP-DCKF-LSTM算法的均方根误差(RMSE)和平均绝对误差(MAE)分别小于0.5%和0.4%,具有较高的SOC估算精度和鲁棒性。