Showing papers on "Kernel adaptive filter published in 2004"

Particle swarm optimization for adaptive IIR filter structures

Abstract: This paper introduces the application of particle swarm optimization techniques to infinite impulse response (IIR) adaptive filter structures. Particle swarm optimization (PSO) is similar to the genetic algorithm (GA) in that it performs a structured randomized search of an unknown parameter space by manipulating a population of parameter estimates to converge on a suitable solution. Unlike the genetic algorithm, particle swarm optimization has not emerged in adaptive filtering literature. Both techniques are independent of the adaptive filter structure and are capable of converging on the global solution for multimodal optimization problems, which makes them especially useful for optimizing IIR and nonlinear adaptive filters. This paper outlines PSO and provides a comparison to the GA for IIR filter structures.

Channel estimation by adaptive interpolation

Abstract: Methods and apparatus that achieve good channel estimation without using unnecessarily complex interpolation filters are described. Adaptive interpolation filtering of a signal in a receiver includes determining at least one correlation function parameter of the channel and determining a filter configuration based on the correlation function parameter. The interpolation may be performed in time, where a Doppler frequency shift can serve as the correlation function parameter, or in frequency, where a root mean square or maximum delay spread can serve as the correlation function parameter, or both. A worst case signal-to-noise ratio may be used in determining the filter configuration, or, optionally, the signal-to-noise ratio can be determined in real time. The filter configuration can be determined in real time or selected from one of a plurality of predetermined configurations having different complexities.

Probability hypothesis density filter versus multiple hypothesis tracking

Abstract: The probability hypothesis density (PHD) filter is a practical alternative to the optimal Bayesian multi-target filter based on finite set statistics. It propagates only the first order moment instead of the full multi-target posterior. Recently, a sequential Monte Carlo (SMC) implementation of PHD filter has been used in multi-target filtering with promising results. In this paper, we will compare the performance of the PHD filter with that of the multiple hypothesis tracking (MHT) that has been widely used in multi-target filtering over the past decades. The Wasserstein distance is used as a measure of the multi-target miss distance in these comparisons. Furthermore, since the PHD filter does not produce target tracks, for comparison purposes, we investigated ways of integrating the data-association functionality into the PHD filter. This has lead us to devise methods for integrating the PHD filter and the MHT filter for target tracking which exploits the advantage of both approaches.

Method and apparatus for improved high-speed adaptive equalization

Abstract: Improved high-speed adaptive equalization is presented that may involve converting an optical signal into an electrical signal and performing equalization by (i) filtering the electrical signal with an analog filter according to at least one filter coefficient to produce a filtered output, (ii) generating an error signal from the filtered output according to an error function, (iii) providing at least one control signal to the analog filter for adjusting the at least one filter coefficient, (iv) detecting a relationship between a change in the at least one filter coefficient and a change in the error signal, and (v) adjusting the at least one filter coefficient according to the relationship to minimize the error signal. The least one coefficient may comprise a plurality of coefficients, and the relationship may be a gradient estimate having multiple components, each determined by varying only one of the coefficients and detecting a resulting change in the error signal.

A variable leaky LMS adaptive algorithm

Abstract: The LMS algorithm has found wide application in many areas of adaptive signal processing and control. We introduce a variable leaky LMS algorithm, designed to overcome the slow convergence of standard LMS in cases of high input eigenvalue spread. The algorithm uses a greedy punish/reward heuristic together with a quantized leak adjustment function to vary the leak. Simulation results confirm that the new algorithm can significantly outperform standard LMS when the input eigenvalue spread is high.

Adaptive two-pass rank order filter to remove impulse noise in highly corrupted images

Abstract: In this paper, we present an adaptive two-pass rank order filter to remove impulse noise in highly corrupted images. When the noise ratio is high, rank order filters, such as the median filter for example, can produce unsatisfactory results. Better results can be obtained by applying the filter twice, which we call two-pass filtering. To further improve the performance, we develop an adaptive two-pass rank order filter. Between the passes of filtering, an adaptive process is used to detect irregularities in the spatial distribution of the estimated impulse noise. The adaptive process then selectively replaces some pixels changed by the first pass of filtering with their original observed pixel values. These pixels are then kept unchanged during the second filtering. In combination, the adaptive process and the second filter eliminate more impulse noise and restore some pixels that are mistakenly altered by the first filtering. As a final result, the reconstructed image maintains a higher degree of fidelity and has a smaller amount of noise. The idea of adaptive two-pass processing can be applied to many rank order filters, such as a center-weighted median filter (CWMF), adaptive CWMF, lower-upper-middle filter, and soft-decision rank-order-mean filter. Results from computer simulations are used to demonstrate the performance of this type of adaptation using a number of basic rank order filters.

Generalized selection weighted vector filters

Abstract: This paper introduces a class of nonlinear multichannel filters capable of removing impulsive noise in color images. The here-proposed generalized selection weighted vector filter class constitutes a powerful filtering framework for multichannel signal processing. Previously defined multichannel filters such as vector median filter, basic vector directional filter, directional-distance filter, weighted vector median filters, and weighted vector directional filters are treated from a global viewpoint using the proposed framework. Robust order-statistic concepts and increased degree of freedom in filter design make the proposed method attractive for a variety of applications. Introduced multichannel sigmoidal adaptation of the filter parameters and its modifications allow to accommodate the filter parameters to varying signal and noise statistics. Simulation studies reported in this paper indicate that the proposed filter class is computationally attractive, yields excellent performance, and is able to preserve fine details and color information while efficiently suppressing impulsive noise. This paper is an extended version of the paper by Lukac et al. presented at the 2003 IEEE-EURASIP Workshop on Nonlinear Signal and Image Processing (NSIP '03) in Grado, Italy.

Design of multiplierless, high-performance, wavelet filter banks with image compression applications

Abstract: The JPEG2000 image coding standard employs the biorthogonal 9/7 wavelet for lossy compression. The performance of a hardware implementation of the 9/7 filter bank depends on the accuracy and the efficiency with which the quantized filter coefficients are represented. A high-precision representation ensures compression performance close to the unquantized, infinite precision filter bank, but at the cost of increased hardware resources and processing time. If the filter coefficients are quantized such that the filter bank properties are preserved, then, the degradation in compression performance will be minimal. This paper investigates two filter structures and two "compensating" filter coefficient quantization methods for improving the performance of multiplierless, quantized filter banks. Rather than using an optimization technique to guide the design process, the new methods utilizes the perfect reconstruction requirements of the filter bank. The results indicate that the best method (a cascade structure with compensating zeros) realizes image-compression performance very similar to the unquantized filter case while also achieving a fast, efficient hardware implementation.

Method of converting resolution of video signals and apparatus using the same

Abstract: A method converts a resolution of video signals, the method including: calculating up-sampling and down-sampling ratios based on a resolution of an input video signal and a desired resolution of an output video signal; calculating a number of filter tabs by multiplying the up-sampling and down-sampling ratios by a number of side lobes; calculating first filter coefficients of a same number of the filter tabs by multiplying a window function by a sinc function; calculating final filter coefficients by subtracting a result of a multiplication of a Gaussian function by a window function from the first filter coefficients, and then normalizing the final filter coefficients; and performing filtering in vertical and horizontal directions based on the final filter coefficients by modifying a sampling rate of an input video signal depending on the up-sampling and down-sampling ratios, to obtain clear video images

Active noise cancellation system

Abstract: In an active noise cancellation system having an adaptive filter that outputs a control signal, first and second speakers that emit a canceling signal generated based on the control signal, a microphone that detects an error signal, a correction filter that corrects the base signal by a correction value to generate a reference signal and a filter coefficient updater that successively updates the adaptive filter coefficient based on the error signal and reference signal such that the error signal is minimized, the correction value of the correction filter is set to a sum obtained by adding the transfer characteristic from the first speaker to the microphone, and a product obtained by multiplying the transfer characteristic from the second speaker to the microphone by the prescribed value, thereby enabling to reduce the number of microphones and avoid the increase in parts, the amount of work to provide complicated wiring to the microphones, and the computational load involved in updating the adaptive filter coefficient, while enabling to maintain an area in which noise can be reduced to the same level as that obtained before reducing the number of microphones.

Adaptive Two-Pass Median Filter Based on Support Vector Machines for Image Restoration

Abstract: In this letter, a novel adaptive filter, the adaptive two-pass median (ATM) filter based on support vector machines (SVMs), is proposed to preserve more image details while effectively suppressing impulse noise for image restoration. The proposed filter is composed of a noise decision maker and two-pass median filters. Our new approach basically uses an SVM impulse detector to judge whether the input pixel is noise. If a pixel is detected as a corrupted pixel, the noise-free reduction median filter will be triggered to replace it. Otherwise, it remains unchanged. Then, to improve the quality of the restored image, a decision impulse filter is put to work in the second-pass filtering procedure. As for the noise suppressing both fixed-valued and random-valued impulses without degrading the quality of the fine details, the results of our extensive experiments demonstrate that the proposed filter outperforms earlier median-based filters in the literature. Our new filter also provides excellent robustness at various percentages of impulse noise.

An unscented particle filter for GMTI tracking

Abstract: Ground moving target indicator (GMTI) tracking is often carried out using extended Kalman filters, as in the variable-structure interacting multiple-model (VS-IMM) filter. In some scenarios, however, this is considered to be inadequate. It has been shown that in this case, a particle filter can give better performance. Such a filter, the variable-structure multiple-model particle filter (VS-MMPF), is given in the literature. In this paper we present a new approach to solving the GMTI tracking problem using a particle filter. We have developed an unscented particle filter, where the particles model the uncertainty over the motion model while, conditional upon the model, the target state is modelled using an unscented Kalman filter. Simulation results show that the UPF-based filter gives performance similar to the VS-MMPF with significantly fewer particles and better results than the standard VS-IMM approach.

λ-WMAP: a statistical speckle filter operating in the wavelet domain

Abstract: In this paper we introduce the Γ-WMAP filter, a wavelet based equivalent to the classical Γ-MAP filter. We model speckle as additive signal-dependent noise, and propose to use the normal inverse Gaussian (NIG) distribution as a statistical model for the wavelet coefficients of both the reflectance image and the noise image. A method for estimating the parameters of the proposed statistical models is presented, and we show that the NIG distribution makes excellent fits to the distributions of the wavelet coefficients of single-look synthetic aperture radar (SAR) images. The performance of the Γ-WMAP filter is tested on three single-look SAR images. We find that when the filter is used in a global mode it may severely blur the image. However, when applied in a local, adaptive mode the new algorithm has excellent de-speckling performance. Visual comparisons with the Γ-MAP filter show that Γ-WMAP tends to give better de-speckling. Quantitative comparisons in homogeneous regions using both the equivalent numbe...

Motion and detail adaptive denoising of video

Abstract: Non-linear techniques for denoising images and video are known to be superior to linear ones. In addition video denoising using spatio-temporal information is considered to be more efficient compared with the use of just temporal information in the presence of fast motion and low noise. Earlier, we introduced a 3-D extension of the K-nearest neighbor filter and have investigated its properties. In this paper we propose a new, motion- and detail-adaptive filter, which solves some of the potential drawbacks of the non-adaptive version: motion caused artifacts and the loss of fine details and texture. We also introduce a novel noise level estimation technique for automatic tuning of the noise-level dependent parameters. The results show that the adaptive K-nearest neighbor filter outperforms the none-adaptive one, as well as some other state-of-the-art spatio-temporal filters such as the 3D alpha-trimmed mean and the state-of-the-art rational filter by Ramponi from both a PSNR and visual quality point of view.

Computation of Asymmetric Signal Extraction Filters and Mean Squared Error for ARIMA Component Models

Abstract: Standard signal extraction results for both stationary and nonstationary time series are expressed as linear filters applied to the observed series. Computation of the filter weights, and of the corresponding frequency response function, is relevant for studying properties of the filter and of the resulting signal extraction estimates. Methods for doing such computations for symmetric, doubly infinite filters are well established. This study develops an algorithm for computing filter weights for asymmetric, semi-infinite signal extraction filters, including the important case of the concurrent filter (for signal extraction at the current time point). The setting is where the time series components being estimated follow autoregressive integrated moving-average (ARIMA) models. The algorithm provides expressions for the asymmetric signal extraction filters as rational polynomial functions of the backshift operator. The filter weights are then readily generated by simple expansion of these expressions, and the corresponding frequency response function is directly evaluated. Recursive expressions are also developed that relate the weights for filters that use successively increasing amounts of data. The results for the filter weights are then used to develop methods for computing mean squared error results for the asymmetric signal extraction estimates.

Method of adaptive interactive learning control and a lithographic manufacturing process and apparatus employing such a method

Abstract: By applying time-frequency analysis to a given standard iterative learning control or ILC an adaptive filter for the learned feed-forward loop is designed. This time varying filter varies according to the momentary frequency content of the error signal and allows to discriminate between areas of deterministic and stochastic error. Its application results in selective application of ILC to those intervals where error signals of high level are concentrated and allows application of a single ILC acquisition to different setpoint trajectories. The adaptive filter finds particular use in lithographic scanning systems where it is used for varying scan length.

De-blocking filter processing apparatus and de-blocking filter processing method

Abstract: A de-blocking filter processing apparatus that achieves high picture quality without consuming processing apparatus power unnecessarily. A loop filter 170 used as a de-blocking filter processing apparatus first acquires a variable-size motion estimation block in a frame for which motion estimation processing is performed. Then, de-blocking filter processing is applied adaptively to a frame for which motion estimation processing is performed, in accordance with the acquired motion estimation block. Application of de-blocking filter processing is executed only at a boundary between a particular motion estimation block and a motion estimation block adjacent to that motion estimation block in a frame for which motion estimation processing is performed.

Computer-aided tuning and diagnosis of microwave filters using sequential parameter extraction

Abstract: This paper describes an approach to the computer-aided tuning of multiple-coupled resonator filters. The method is based on a sequential parameter extraction and tuning process and employs three different filter models. A detuned model represents the initial status of the filter after a well defined detuning procedure. The target filter is described by the ideal model, whereas the actual state of the filter at each tuning step is represented by a coarse adaptive model. The goal of the procedure is the convergence of the coarse model to the ideal model and will be obtained by systematically centering resonant frequencies and coupling coefficients. Two practical examples comprising a low- and a high-degree filter confirm the effectiveness of the proposed approach.

Modified-filtered-x LMS algorithm based active noise control systems with improved online secondary-path modeling

Abstract: This paper proposes a new method for active noise control systems with online secondary-path modeling. The main features of the proposed method are that the modified-FxLMS (MFxLMS) algorithm is used in adapting the noise control filter and a new variable step-size (VSS) LMS algorithm is used in the secondary-path modeling filter. The step-size is varied in accordance with the power of the residual error signal (the desired response for the modeling filter). It is found that the desired response for the modeling filter is corrupted by a noise which is decreasing in nature, (ideally) converging to zero. Hence, a small step-size is used initially and later its value is increased accordingly. Computer simulations show that the proposed method gives better performance than the existing methods.

Sub-band adaptive filtering with delay compensation for active control

Abstract: This paper introduces a filtered-x least-mean-square (FXLMS) based sub-band adaptive algorithm with error path delay compensation. Our algorithm avoids the signal path delay while compensating for the error path delay, thus increasing system stability. Simulation results are presented to demonstrate experimentally the efficiency of this new adaptive algorithm.

Covariance profiling for an adaptive Kalman filter to suppress sensor quantization effects

Abstract: This paper presents a generic approach to model the noise covariance associated with discrete sensors such as incremental encoders and low resolution analog to digital converters. The covariance is then used in an adaptive Kalman filter that selectively and appropriately carries out measurement updates. The temporal as well as system state measurements are used to predict the quantization error of the measurement signal. The effectiveness of the method is demonstrated by applying the technique to incremental encoders of varying resolutions. Simulation of an example system with varying encoder resolutions is presented to show the performance of the new filter. Results show that the new adaptive filter produces more accurate results while requiring a lower resolution encoder than a similarly designed conventional Kalman filter, especially at low velocities.

Voice activity detection with adaptive noise floor tracking

Abstract: The present invention relates to a method and apparatus for detecting voice activity in a communication signal, wherein filter means are provided for estimating or suppressing an offset component of the level of the communication signal. A filter parameter is controlled based on the output of the filter means. Furthermore, the estimation or suppression of the offset component is limited in response to the output of the filter means. The filter means may be based on a non-linear adaptive notch level filter or a noise floor tracking filter. Thereby, the tracking behavior of noise floor estimation to sudden rises in noise floor can be improved and the voice activity detection can work efficiently over a wide dynamic range.

Seek trajectory generation using adaptive filtering

Abstract: Embodiments of the present invention provide a method and a system for generating optimal feedforward signal for the seek control to suppress the RTV (Random Transient Vibrations) and the seek acoustic noise. One aspect is directed to a method of providing a revised feedforward signal using an adaptive filter in a feedforward control system for controlling an actuator to move a head to seek a track and settle on the track of a disk in a disk drive apparatus. The method comprises performing a seek operation of the head using an initial feedforward signal; obtaining an error signal at settling after performing the seek operation; determining filter characteristics of the adaptive filter to minimize the error signal; and implementing the adaptive filter having the determined filter characteristics in the feedforward control system to produce a revised feedforward signal for controlling the actuator for moving the head in the disk drive apparatus.

Stochastic mean-square performance analysis of an adaptive Hammerstein filter

Abstract: This paper presents an almost sure (a.s.) mean-square performance analysis of an adaptive Hammerstein filter for the case when the measurement noise in the desired response signal is a martingale difference sequence. The system model consists of a series connection of a memoryless nonlinearity followed by a recursive linear filter. It is shown under the conditions of the analysis that the long-term time average of the squared excess estimation error of the adaptive filter can be made arbitrarily close to zero.

Parallel Filter Algorithms for Data Assimilation in Oceanography

Abstract: A consistent systematic comparison of filter algorithms based on the Kalman filter and intended for data assimilation with large-scale nonlinear models is presented. Considered are the EnsembleKalman Filter (EnKF), the Singular Evolutive Extended Kalman (SEEK) filter, and the Singular Evolutive Interpolated Kalman (SEIK) filter. Within the two parts of this thesis, the filter algorithms are compared with a focus on their mathematical properties as Error Subspace KalmanFilters (ESKF). Further, the filters are studied as parallel algorithms. This study includes the development of an efficient framework for parallel filtering. In the first part, the filters are motivated in the context of statistical estimation. The unified interpretation as ESKF algorithms provides the basis for the consistent comparison of the filters. Numerical data assimilation experiments with a model based on the shallow water equations show how choices of the filter schemeand particular state ensembles for the filter initialization lead to variations of the data assimilation performance.The application of the three filter algorithms on parallel computers is studied in the second part. The parallelization possibilities of the different phases of the algorithms are examined. Further, a framework for parallel filtering is developed which allows to combine filter algorithms with existing numerical models requiring only minimal changes to the source code of the model.The framework is used to combine the parallel filters with the 3D finite element ocean model FEOM. Numerical data assimilation experiments are utilized to assess the parallel efficiency of the filtering framework and the parallel filters. The experiments yield an excellent parallel efficiency for the filtering framework. Further, the framework and the filter algorithms are well suited for application to realistic large-scale data assimilation problems.

Using difference kernels for image filtering

Abstract: Methods and apparatus, including computer program products, for filtering an image. A filter kernel is received to determine one or more filtered values for each pixel in a sequence of pixels, where adjacent pixels are separated by a characteristic distance in the image. A difference kernel is defined based on local differences between a first kernel and a second kernel that are defined by the filter kernel centered at a first location and a second location, respectively. The second location is separated from the first location by the characteristic distance separating adjacent pixels in the sequence. The difference kernel is used to determine a difference between filtered values of adjacent pixels in the sequence. For depth of field filtering, the filter kernel can include a blur filter kernel that is based upon depth values of pixels in the sequence.