Showing papers on "Kernel adaptive filter published in 2001"

Noise adaptive soft-switching median filter

Abstract: Existing state-of-the-art switching-based median filters are commonly found to be nonadaptive to noise density variations and prone to misclassifying pixel characteristics at high noise density interference. This reveals the critical need of having a sophisticated switching scheme and an adaptive weighted median filter. We propose a novel switching-based median filter with incorporation of fuzzy-set concept, called the noise adaptive soft-switching median (NASM) filter, to achieve much improved filtering performance in terms of effectiveness in removing impulse noise while preserving signal details and robustness in combating noise density variations. The proposed NASM filter consists of two stages. A soft-switching noise-detection scheme is developed to classify each pixel to be uncorrupted pixel, isolated impulse noise, nonisolated impulse noise or image object's edge pixel. "No filtering" (or identity filter), standard median (SM) filter or our developed fuzzy weighted median (FWM) filter will then be employed according to the respective characteristic type identified. Experimental results show that our NASM filter impressively outperforms other techniques by achieving fairly close performance to that of ideal-switching median filter across a wide range of noise densities, ranging from 10% to 70%.

Stochastic Methods for Sequential Data Assimilation in Strongly Nonlinear Systems

Abstract: This paper considers several filtering methods of stochastic nature, based on Monte Carlo drawing, for the sequential data assimilation in nonlinear models. They include some known methods such as the particle filter and the ensemble Kalman filter and some others introduced by the author: the second-order ensemble Kalman filter and the singular extended interpolated filter. The aim is to study their behavior in the simple nonlinear chaotic Lorenz system, in the hope of getting some insight into more complex models. It is seen that these filters perform satisfactory, but the new filters introduced have the advantage of being less costly. This is achieved through the concept of second-order-exact drawing and the selective error correction, parallel to the tangent space of the attractor of the system (which is of low dimension). Also introduced is the use of a forgetting factor, which could enhance significantly the filter stability in this nonlinear context.

Space variant median filters for the restoration of impulse noise corrupted images

Abstract: This brief proposes a generalized framework of median based switching schemes, called multi-state median (MSM) filter. By using a simple thresholding logic, the output of the MSM filter is adaptively switched among those of a group of center weighted median (CWM) filters that have different center weights. As a result, the MSM filter is equivalent to an adaptive CWM filter with a space varying center weight which is dependent on local signal statistics. The efficacy of the proposed filter has been evaluated by extensive simulations.

Filtering of laser altimetry data using a slope adaptive filter

Abstract: A point set obtained by laser altimetry represents points from not only the ground surface but also objects found on it. For civil works applications points representing the surface of non-ground objects have to be removed from the point set in a filtering process. This paper describes modifications made to an existing “slope based” filtering algorithm, and presents some results obtained from the use of the filter. The “slope based” filter operates on the assumption that terrain slopes do not rise above a certain threshold, and that features in the data that have slopes above this threshold do not belong to the natural terrain surface. However, this assumption limits the use of the filter to terrain with gentle slopes. To overcome this limitation, the filter was modified in manner that the threshold varies with respect to the slope of the terrain. The results of tests carried out using the modified filter confirm that the modification reduces the number of Type I errors (ground points in steep terrain are not filtered off). Further numerical comparison of the filter output with a reference data set for the same site (obtained photogrammetrically) show that the filter generates relatively minimal Type II errors. The output of the modified slope filter was also compared with the output from a filtering found in the commercial software package, “Terrascan”.

Filtering of multichannel SAR images

Abstract: An explicit form of the linear multichannel synthetic aperture radar (SAR) intensity filter, which preserves radiometry while optimally reducing speckle is derived, together with a compact expression for the theoretical gain in equivalent numbers of looks (ENLs). The filter can be applied to mixed data types, which is demonstrated using a combination of ERS and JERS satellite data, and confirms the filter performance predicted by the theory. Tests indicate that a simplified form of the filter, which neglects correlation between images, gives an ENL only slightly less than optimal, while being much easier to implement. Exact analysis of the effect of estimating filter weights shows that the linear increase in ENL with the number of images predicted for the ideal filter does not occur. In practice, the ENL is affected by the window size used to estimate the weights and saturates as the number of images increases. An efficient recursive form of the filter is described, which is most naturally applied to multitemporal data for the practically important case where the current image is uncorrelated with previous images in a data sequence.

Fast impulsive noise removal

Abstract: A generic n-dimensional filter with the primary purpose of eliminating impulsive-like noise is presented. This recursive nonlinear filter is composed of two conditional rules, which are applied independently, in any order, one after the other. It identifies noisy items by inspection of their surrounding neighborhood, and afterwards it replaces their values with the most "conservative" ones out of their neighbors' values. In this way, no new values are introduced and the histogram distribution range is conserved. This n-dimensional filter can be decomposed recursively to a lower dimensional space, each time generating two sets of n(n-1)-dimensional filters. This study, which focuses on the case of two-dimensional signals (gray scale images), explores one possible implementation of this new filter and orients the evaluation of its performance toward the median filter, as this filter is the basis of many more sophisticated filters for impulsive noise reduction. Tests were carried out using both real and artificial images. We found this new filter to be much faster than the median filter while performing comparably in terms of both image information conservation and noise reduction, which suggests that it could replace the median filter for the preliminary processing included in state-of-the-art noise removal filters. This new filter should either eliminate or attenuate most noisy pixels in synthetic and natural images not excessively contaminated. It has a slight smoothing effect on nonnoisy image regions. In addition, it is scalable, easily implemented, and adaptable to specific applications.

Adaptive filtering algorithms with selective partial updates

Abstract: In some applications of adaptive filtering such as active noise reduction, and network and acoustic echo cancellation, the adaptive filter may be required to have a large number of coefficients in order to model the unknown physical medium with sufficient accuracy. The computational complexity of adaptation algorithms is proportional to the number of filter coefficients. This implies that, for long adaptive filters, the adaptation task can become prohibitively expensive, ruling out cost-effective implementation on digital signal processors. The purpose of partial coefficient updates is to reduce the computational complexity of an adaptive filter by adapting a block of the filter coefficients rather than the entire filter at every iteration. In this paper, we develop a selective-partial-update normalized least-mean-square (NLMS) algorithm, and analyze its stability using the traditional independence assumptions and error-energy bounds. Selective partial updating is also extended to the affine projection (AP) algorithm by introducing multiple constraints. The new algorithms appear to have good convergence performance as attested to by computer simulations with real speech signals.

Automated filter tuning using generalized low-pass prototype networks and gradient-based parameter extraction

Abstract: A novel technique for automated filter tuning is introduced. The filter to be tuned is represented by a generalized filter low-pass prototype model rather than a specialized equivalent network. The prototype model is based on the minimum number of characteristic filter parameters to represent the filter transfer function correctly. The parameter values are found from a gradient-based parameter-extraction process using measured S-parameters. Automated filter tuning is performed as a two-step procedure. First, the parameter sensitivities with respect to the tuning elements are determined by a series of S-parameter measurements. Second, the parameter values of the filter are compared to the values of the ideal filter prototype found from a filter synthesis, thus yielding the optimal screw positions. This novel tuning technique has been tested successfully with direct coupled three-resonator and cross-coupled four- and six-resonator filters.

Multi-channel signal encoding and decoding

Abstract: A multi-channel linear predictive analysis-by-synthesis signal encoding method detects (S26, S27) inter-channel correlation and select one of several possible encoding modes (S24, S29, S30) based on the detected correlation.

Delayed and asequent data in decentralized sensing networks

Abstract: This paper presents an exact solution to the delayed data problem for the information form of the Kalman filter, together with its application to decentralised sensing networks. To date, the most common method of handling delayed data in sensing networks has been to use a conservative time alignment of the observation data with the filter time. However, by accounting for the correlation between the late data and the filter over the delayed period, an exact solution is possible. The inclusion of this information correlation term adds little extra complexity, and may be applied in an information filter update stage which is associative. The delayed data algorithm can also be used to handle data that is asequent or out of order. The asequent data problem is presented in a simple recursive information filter form. The information filter equations presented in this paper are applied in a decentralised picture compilation problem. This involves multiple aircraft tracking multiple ground targets and the construction of a single common tactical picture.

An efficient systolic architecture for the DLMS adaptive filter and its applications

Abstract: In this paper, we propose an efficient systolic architecture for the delay least-mean-square (DLMS) adaptive finite impulse response (FIR) digital filter based on a new tree-systolic processing element (PE) and an optimized tree-level rule. Applying our tree-systolic PE, a higher convergence rate than that of the conventional DLMS structures can be obtained without sacrificing the properties of the systolic-array architecture. The efficient systolic adaptive FIR digital filter not only operates at the highest throughput in the word-level but also considers finite driving/update of the feedback error signal. Furthermore, based on our proposed optimized tree-level rule that takes account of minimum delay and high regularity, an efficient N-tap systolic adaptive FIR digital filter can be easily determined under the constraint of maximum driving of the feedback error signal.

A Simple Neural Network Pruning Algorithm with Application to Filter Synthesis

Abstract: This paper describes an approach to synthesizing desired filters using a multilayer neural network (NN). In order to acquire the right function of the object filter, a simple method for reducing the structures of both the input and the hidden layers of the NN is proposed. In the proposed method, the units are removed from the NN on the basis of the influence of removing each unit on the error, and the NN is retrained to recover the damage of the removal. Each process is performed alternately, and then the structure is reduced. Experiments to synthesize a known filter were performed. By the analysis of the NN obtained by the proposed method, it has been shown that it acquires the right function of the object filter. By the experiment to synthesize the filter for solving real signal processing tasks, it has been shown that the NN obtained by the proposed method is superior to that obtained by the conventional method in terms of the filter performance and the computational cost.

Adaptive digital filters

Abstract: This text emphasizes the intricate relationship between adaptive filtering and signal analysis - highlighting stochastic processes, signal representations and properties, analytical tools, and implementation methods. This second edition includes new chapters on adaptive techniques in communications and rotation-based algorithms. It provides practical applications in information, estimation, and circuit theories.

A WISE method for designing IIR filters

Abstract: The problem of designing optimal digital IIR filters with frequency responses approximating arbitrarily chosen complex functions is considered. The real-valued coefficients of the filter's transfer function are obtained by numerical minimization of carefully formulated cost, which is referred here to as the weighted integral of the squared error (WISE) criterion. The WISE criterion linearly combines the WLS criterion that is used in the weighted least squares approach toward filter design and some time-domain components. The WLS part of WISE enforces the quality of the frequency response of the designed filter, while the time-domain part of the WISE criterion restricts the positions of the filter's poles to the interior of an origin-centred circle with arbitrary radius. This allows one not only to achieve stability of the filter but also to maintain some safety margins. A great advantage of the proposed approach is that it does not impose any constraints on the optimization problem and the optimal filter can be sought using off-the-shelf optimization procedures. The power of the proposed approach is illustrated with filter design examples that compare favorably with results published in research literature.

A normalized gradient descent algorithm for nonlinear adaptive filters using a gradient adaptive step size

Abstract: A fully adaptive normalized nonlinear gradient descent (FANNGD) algorithm for online adaptation of nonlinear neural filters is proposed. An adaptive stepsize that minimizes the instantaneous output error of the filter is derived using a linearization performed by a Taylor series expansion of the output error. For rigor, the remainder of the truncated Taylor series expansion within the expression for the adaptive learning rate is made adaptive and is updated using gradient descent. The FANNGD algorithm is shown to converge faster than previously introduced algorithms of this kind.

Efficient implementation of all-digital interpolation

Abstract: B-splines are commonly used for continuous representation of discrete time signals. This kind of representation proves to be very useful in applications such as image interpolation, rotation and edge detection. In all these applications, the first step is to compute the B-spline coefficients of the signal, and this involves the use of an IIR noncausal filter called the direct B-spline filter. The signal reconstruction is achieved using the indirect B-spline filter, which in many applications operates at a higher rate. We introduce a simplified implementation of the signal reconstruction part that will significantly reduce the overall complexity. We also show that the direct B-spline filter can safely be replaced with a short FIR filter, without compromising the performance of the traditional method. Numerous examples show both visually and numerically that the differences between this method and the traditional one are indeed very small. Finally, we report the performance of these newly proposed methods in other image processing applications such as edge detection and least squares approximation.

The interacting multiple model algorithm for accurate state estimation of maneuvering targets

Abstract: Accurate state estimation of targets with changing dynamics can be achieved through the use of multiple filter models. The interacting multiple model (IMM) algorithm provides a structure to efficiently manage multiple filter models. Design of an IMM requires selection of the number and type of filter models and selection of each of the individual filter parameters. In this article the results for five filter models on 10 target trajectory segments are discussed and compared. The complexity of the filter models increases from a single constant velocity model to a three-model IMM filter. The results show that the overall performance of the state estimates, for most targets, improves as the complexity of the filter models increases. Selection of IMM filter parameters is addressed and results are provided to show that performance of the IMM appears to be relatively insensitive to large changes in filter parameters. The performance of an IMM is primarily determined by the selection of the component filter models.

Robust M-estimate adaptive filtering

Abstract: An M-estimate adaptive filter for robust adaptive filtering in impulse noise is proposed. Instead of using the conventional least-square cost function, a new cost function based on an M-estimator is used to suppress the effect of impulse noise on the filter weights. The resulting optimal weight vector is governed by an M-estimate normal equation. A recursive least M-estimate (RLM) adaptive algorithm and a robust threshold estimation method are derived for solving this equation. The mean convergence performance of the proposed algorithm is also analysed using the modified Huber function (a simple but good approximation to the Hampel's three-parts-redescending M-estimate function) and the contaminated gaussian noise model. Simulation results show that the proposed RLM algorithm has better performance than other recursive least squares (RLS) like algorithms under either contaminated gaussian or alpha-stable noise environment. The initial convergence, steady-state error, robustness to system change and computational complexity are also found to be comparable to the conventional RLS algorithm under gaussian noise alone.

Adaptive Sine-Modulated/Cosine-Modulated Filter Bank Equalizer for Transmultiplexers

Abstract: Cosine-modulated filter banks provide an efficient realization for filter bank-based trans- multiplexers. This paper describes an equalization method for these systems. At the receiver we have two parallel analysis filter banks (cosine- and sine- modulated FBs) whose subchannel outputs are com- bined optimally. This increases the computational burden on the filter bank side, but very simple sub- channel equalizer structure is a clear advantage. For adaptive implementation, the number of required general multipliers is reduced, and it is possible to develop LMS-based equalization algorithms with fast convergence. Therefore, the equalizer structure is called as Adaptive SMFB/CMFB Equalizer for Transmultiplexers (ASCET).

Sliding-window transform with integrated windowing

Abstract: A system for a sliding-window transform with integrated windowing is described. The system provides a Direct Fourier Transform kernel with an integrated windowing filter having a desired number of stages. In one embodiment, the windowing filter is a lowpass filter. In one embodiment, the lowpass filter has a rectangular filter transfer characteristic. The DFT includes a complex multiplier. A first portion of the windowing filter is provided before the complex multiplier and can be implemented using real arithmetic. A second portion of the windowing filter is provided after the complex multiplier and is implemented using complex arithmetic. In one embodiment, the filter weights of the second portion of the windowing filter are unity and thus no multiplier is needed for the filter weights in the second portion of the windowing filter.

Spatio-temporal joint filter for noise reduction

Abstract: A spatio-temporal joint filter and a spatial joint filter for noise reduction are disclosed. The spatio-temporal joint filter includes a spatial joint filter including the first and second sub filters having different characteristics and includes a temporal joint filter. When the present invention is adequately used, an edge/detail region of an image is well preserved, an aggressive noise reduction is performed on a flat region, and the temporal flicker problems are eliminated. Additionally, it has an intrinsic motion compensation effect by using the spatio-temporal correlation between the adjacent frames.

Method and apparatus for providing blind adaptive estimation and reception

Abstract: A method is disclosed to minimize the mean-square-error of an estimate of an unknown parameter, such as a data symbol transmitted through a channel, such as a WCDMA channel. The method includes steps of (a) replacing a required multiplication of an input signal vector by an inverse covariance matrix, which is one of a total signal covariance matrix or an interference-plus-noise covariance matrix, by linear filtering, wherein directly computed or estimated filter elements of a row or a column of the inverse covariance matrix, corresponding to time instant i, are used as linear filter coefficients; (b) forming a vector g(i) from the filter outputs, the vector g(i) being estimated element by element using the linear filter; and (c) using the vector g(i) in place of a vector that would have been obtained by directly multiplying the signal vector by the inverse covariance matrix. The linear filter w(i) converges or closely converges to a row or column of the required inverse covariance matrix. For the case of the interference-plus-noise covariance matrix, when a desired filter is found through adaptation, the desired filter is used to filter the total signal, including the desired signals, to generate the output vector g(i). The filter outputs generated using either of the total signal covariance matrix or the interference-plus-noise covariance matrix are employed to estimate an unknown parameter through the use of a further filter matched to the pulse shape of the unknown parameter carried by the input signal, where the further filter is a RAKE receiver that performs multipath combining at a chip level and single code correlation, or a RAKE receiver that performs despreading using a code correlator bank, and multipath combining at the symbol level using the correlator outputs.

Generating resized images using ripple free image filtering

Abstract: An optimal filter kernel, formed by convolving a box filter with a filter of fixed integer width and unity area, is used to perform image resizing and reconstruction. The optimal filter has forced zeros at locations along a frequency scale corresponding to the reciprocal of the spacing of one or more pixels that comprise a source image to be resized. When a rescale value for a source image is selected, the optimal filter kernel is computed, mapped to the source image, and centered upon a location within the source image corresponding to the position of an output pixel to be generated. The number of pixels that lie underneath the optimal filter kernel is established by multiplying the number of pixels that comprise the width of the source image by the selected rescale value. Upon mapping the optimal filter kernel, the output pixel values that comprise the resized image are then evaluated by processing the one or more source image pixels, such as through interpolation. Alternatively, the output pixel values of the resized image are calculated by performing partial integral analysis with respect to a standard filter kernel of fixed width and unity area. The output pixel values are calculated by multiplying the pixel value for each pixel under the kernel by the area of the standard filter kernel surrounding the pixel. The products are then summed to reveal the output pixel value, and placed into the output image buffer. Both of these methods speed up the computation process, while producing a ripple free output image.

Adaptive spatio-temporal filter for human vision system models

Abstract: An adaptive spatio-temporal filter for use in video quality of service instruments based on human vision system models has a pair of parallel, lowpass, spatio-temporal filters receiving a common video input signal. The outputs from the pair of lowpass spatio-temporal filters are differenced to produce the output of the adaptive spatio-temporal filter, with the bandwidths of the pair being such as to produce an overall bandpass response. A filter adaptation controller generates adaptive filter coefficients for each pixel processed based on a perceptual parameter, such as the local average luminance, contrast, etc., of either the input video signal or the output of one of the pair of lowpass spatio-temporal filters. Each of the pair of lowpass spatio-temporal filters has a temporal IIR filter in cascade with a 2-D spatial IIR filter, and each individual filter is composed of a common building block,5 i.e., a first order, unity DC gain, tunable lowpass filter having a topology suitable for IC implementation. At least two of the building blocks make up each filter with the overall adaptive spatio-temporal filter response having a linear portion and a non-linear portion, the linear portion being dominant at low luminance levels and the non-linear portion being consistent with enhanced perceived brightness as the luminance level increases.

Partitioned block frequency domain adaptive filter

Abstract: The partitioned block frequency domain adaptive filter according to the invention comprises a plurality of parallel arranged filter partitions. Each filter partition models a part of an impulse response of the adaptive filter and has update means for updating filter coefficients of that filter partition by means of a circular convolution. The update means intermittently constrain these filter coefficients by eliminating circular wrap-around artifacts of the circular convolution. The update means are arranged for updating the filter coefficients in dependence on at least part of the circular wrap-around artifacts of adjacent update means, resulting in an improved convergence behavior of the adaptive filter.

Noise reduction method and device

Abstract: The invention concerns a method which consists, when analysing an input signal in the frequency domain, in determining a noise level estimator and a useful signal level estimator in an input signal frame, thereby enabling to calculate the transfer function of a first noise-reducing filter, carrying out a second pass to fine-tune the useful signal level estimator, by combining the signal spectrum and the first filter transfer function, then to calculate the transfer function of a second noise-reducing filter on the basis of the fine-tuned useful signal level estimator and the noise level estimator. Said second noise-reducing filter is then used to reduce the noise level in the frame.

Stochastic analysis of the LMS algorithm with a saturation nonlinearity following the adaptive filter output

Abstract: This paper presents a statistical analysis of the least mean square (LMS) algorithm with a zero-memory scaled error function nonlinearity following the adaptive filter output. This structure models saturation effects in active noise and active vibration control systems when the acoustic transducers are driven by large amplitude signals. The problem is first defined as a nonlinear signal estimation problem and the mean-square error (MSE) performance surface is studied. Analytical expressions are obtained for the optimum weight vector and the minimum achievable MSE as functions of the saturation. These results are useful for adaptive algorithm design and evaluation. The LMS algorithm behavior with saturation is analyzed for Gaussian inputs and slow adaptation. Deterministic nonlinear recursions are obtained for the time-varying mean weight and MSE behavior. Simplified results are derived for white inputs and small step sizes. Monte Carlo simulations display excellent agreement with the theoretical predictions, even for relatively large step sizes. The new analytical results accurately predict the effect of saturation on the LMS adaptive filter behavior.

An efficient approach for designing nearly perfect-reconstruction cosine-modulated and modified DFT filter banks

Abstract: Efficient two-step algorithms are described for optimizing the stopband response of the prototype filter for cosine-modulated and modified DFT filter banks either in the minimax or in the least-mean-square sense subject to the maximum allowable aliasing and amplitude errors. The first step involves finding a good start-up solution using a simple technique. This solution is improved in the second step by using nonlinear optimization. Several examples are included illustrating the flexibility of the proposed approach for making compromises between the required filter lengths and the aliasing and amplitude errors. These examples show that by allowing very small amplitude and aliasing errors, the stopband performance of the resulting filter bank is significantly improved compared to the corresponding perfect-reconstruction filter bank. Alternatively, the filter orders and, consequently, the overall delay can be significantly reduced to achieve practically the same performance.