scispace - formally typeset
Search or ask a question

Showing papers on "Kernel adaptive filter published in 1992"


Journal ArticleDOI
TL;DR: A least-mean-square adaptive filter with a variable step size, allowing the adaptive filter to track changes in the system as well as produce a small steady state error, is introduced.
Abstract: A least-mean-square (LMS) adaptive filter with a variable step size is introduced. The step size increases or decreases as the mean-square error increases or decreases, allowing the adaptive filter to track changes in the system as well as produce a small steady state error. The convergence and steady-state behavior of the algorithm are analyzed. The results reduce to well-known results when specialized to the constant-step-size case. Simulation results are presented to support the analysis and to compare the performance of the algorithm with the usual LMS algorithm and another variable-step-size algorithm. They show that its performance compares favorably with these existing algorithms. >

966 citations


Journal ArticleDOI
TL;DR: In this article, an exact analysis of the critically subsampled two-band modelization scheme is given, and it is demonstrated that adaptive cross-filters between the subbands are necessary for modelization with small output errors.
Abstract: An exact analysis of the critically subsampled two-band modelization scheme is given, and it is demonstrated that adaptive cross-filters between the subbands are necessary for modelization with small output errors. It is shown that perfect reconstruction filter banks can yield exact modelization. These results are extended to the critically subsampled multiband schemes, and important computational savings are seen to be achieved by using good quality filter banks. The problem of adaptive identification in critically subsampled subbands is considered and an appropriate adaptation algorithm is derived. The authors give a detailed analysis of the computational complexity of all the discussed schemes, and experimentally verify the theoretical results that are obtained. The adaptive behavior of the subband schemes that were tested is discussed. >

552 citations


01 Jan 1992
TL;DR: An exact analysis of the critically subsampled two-band modelization scheme is given, and it is demonstrated that adaptive cross-filters between the subbands are necessary for modelization with small output errors.
Abstract: Adaptive filtering in subbands is a new technique for the real-time identification of large impulse responses like the ones encountered in acoustic echo cancellation. This tech- nique generally allows computational savings as well as better convergence behavior. We give first an exact analysis of the critically subsampled two band modelization scheme. We demonstrate that adaptive cross-filters between the subbands are necessary for modeliza- tion with small output errors; moreover, we show that perfect reconstruction filter banks can yield exact modelization. We ex- tend those results to the critically subsampled multiband schemes, and we show that important computational savings can be achieved by using good quality filter banks. Then we consider the problem of adaptive identification in critically sub- sampled subbands, and we derive an appropriate adaptation algorithm. We give a detailed analysis of the computational complexity of all the discussed schemes, and we verify experi- mentally the theoretical results that we have obtained. Finally, we discuss the adaptive behavior of the subband schemes that we have tested. We generally observe some degradation of the convergence performance in comparison with conventional schemes; however, the overall performance could be acceptable in practical use.

519 citations


Proceedings ArticleDOI
23 Mar 1992
TL;DR: The authors apply the criterion used in the unbiased estimation of log spectrum to the spectral model represented by the mel-cepstral coefficients to solve the nonlinear minimization problem involved in the method and derive an adaptive algorithm whose convergence is guaranteed.
Abstract: The authors describe a mel-cepstral analysis method and its adaptive algorithm. In the proposed method, the authors apply the criterion used in the unbiased estimation of log spectrum to the spectral model represented by the mel-cepstral coefficients. To solve the nonlinear minimization problem involved in the method, they give an iterative algorithm whose convergence is guaranteed. Furthermore, they derive an adaptive algorithm for the mel-cepstral analysis by introducing an instantaneous estimate for gradient of the criterion. The adaptive mel-cepstral analysis system is implemented with an IIR adaptive filter which has an exponential transfer function, and whose stability is guaranteed. The authors also present examples of speech analysis and results of an isolated word recognition experiment. >

374 citations


Journal ArticleDOI
TL;DR: A set of time-domain conditions for reconstruction which can be used directly in a filter bank design procedure is derived, which allows for the design of many useful banks.
Abstract: The authors present a new time-domain approach for the analysis and design of a broad class of general analysis/synthesis systems based on M-band filter banks. They derive a set of time-domain conditions for reconstruction which can be used directly in a filter bank design procedure. The general and unrestricted nature of this framework allows for the design of many useful banks. In addition to the complete derivation of the time-domain conditions, they also describe the associated filter bank design procedure and a number of design examples are included. >

180 citations


Journal ArticleDOI
TL;DR: In this paper, an adaptive impulse correlated filter (AICF) was proposed to estimate the deterministic component of the signal and remove the noise uncorrelated with the stimulus even if this noise is colored, as in the case of evoked potentials.
Abstract: An adaptive impulse correlated filter (AICF) for event-related signals that are time-locked to a stimulus is presented. This filter estimates the deterministic component of the signal and removes the noise uncorrelated with the stimulus, even if this noise is colored, as in the case of evoked potentials. The filter needs two inputs: the signal (primary input) and an impulse correlated with the deterministic component (reference input). The LMS algorithm is used to adjust the weights in the adaptive process. It is shown that the AICF is equivalent to exponentially weighted averaging (FWA) when using the LMS algorithm. A quantitative analysis of the signal-to-noise ratio improvement, convergence, and misadjustment error is presented. A comparison of the AICF with ensemble averaging (EA) and moving window averaging (MWA) techniques is also presented. The adaptive filter is applied to real high-resolution ECG signals and time-varying somatosensory evoked potentials. >

144 citations


Patent
18 Jul 1992
TL;DR: In this article, a method and apparatus for adaptively equalizing data signals in a communications receiver is provided for the recovery of multilevel amplitude modulated data, such as QAM data.
Abstract: A method and apparatus are provided for adaptively equalizing data signals in a communications receiver. An unequalized data signal is demodulated. The demodulated data signal is filtered in an adaptive equalizer (60) that initially updates adaptive filter coefficients using error signals derived from a first algorithm. A carrier lock signal is generated (62, 78) when a phase error of a filtered signal output from the adaptive equalizer reaches a threshold value. The adaptive filter coefficients are updated (74) using error signals derived from a second algorithm instead of the first algorithm in response to the carrier lock signal (72). The first algorithm is a self-recovering equalization algorithm such as the Constant Modulus Algorithm. The second algorithm can be a decision directed algorithm. Carrier phase is recovered without the use of a phase rotator or phase de-rotator, by locating the adaptive equalizer inside of the carrier recovery loop (56). The invention is particularly adapted for use in the recovery of multilevel amplitude modulated data, such as QAM data.

126 citations


Patent
Paul G. Roetling1
12 Aug 1992
TL;DR: In this article, an image processing system is provided to convert halftone images to continuous tone images, which employs an adaptive filter which processes successive pixels in an input Halftone image, and a control operates the adaptive filter to apply one of the predetermined filters to the current pixel as a function of the associated pixel spatial gradient.
Abstract: An image processing system is provided to convert halftone images to continuous tone images. It employs an adaptive filter which processes successive pixels in an input halftone image. The adaptive filter employs a filter that is selected under feedback control from a plurality of filter sets each having a plurality of filters. The halftone image is also low-pass filtered to generate a first approximation image (FAI). A spatial gradient value is computed for each pixel in the FAI. A control operates the adaptive filter to apply one of the predetermined filters to the current pixel as a function of the associated pixel spatial gradient. An output image from the adaptive filter in a first iteration of the filtering procedure can then be applied to the input of the adaptive filter for a second adaptive filtering iteration. Pixel gradients for the second iteration are computed from the image output from the first iteration. A predetermined number of iterations are performed and the image output from the last iteration is a continuous tone image for system output.

115 citations


Patent
02 Nov 1992
TL;DR: In this paper, a weight adjustment unit was proposed for adaptive digital filters, where the weights of an adaptive digital filter were adjusted according to one or more input signals to the digital filter and according to an error signal indicative of the difference between the actual and desired outputs of the digital filters.
Abstract: An adaptive digital filter uses a weight adjustment unit for adjusting the weights of an adaptive digital filter according to one or more input signals to the digital filter and according to an error signal indicative of the difference between the actual and desired outputs of the digital filter. The weight adjustment unit has a first low-pass filter for low-pass filtering a signal indicative of the product of the error signal and the one or more input signals, a squarer for squaring the output of the first low-pass filter, a second low-pass filter for low-pass filtering the output of the squarer to extract the D.C. component thereof, a third low-pass filter for low-pass filtering a signal indicative of the output of the error signal squared to extract the D.C. component thereof, a dividing unit for dividing the output of the second low-pass filter by the output of the third low-pass filter to provide a loop bandwidth, and a weight calculation unit for providing values for one or more weights of the adaptive digital filter according to the previous values of the weights and the value of the loop bandwidth.

90 citations


Journal ArticleDOI
TL;DR: In this paper, a new adaptive filter structure is introduced that permits a closer placement of the transducers and that allows the cancellation of noise in the presence of crosstalk, which shows considerable improvement in mean-square error over that obtained with standard noise canceling algorithms.
Abstract: The application of adaptive filters in noise canceling often requires the relative placement of the two transducers at a distance that necessitates a larger order filter in order to obtain an adequate output signal-to-noise ratio. A new adaptive filter structure is introduced that permits a closer placement of the transducers and that allows the cancellation of noise in the presence of crosstalk. Algorithms are developed for the new transversal and lattice filter estimators. Simulations show considerable improvement in mean-square error over that obtained with standard noise canceling algorithms. >

89 citations


Proceedings ArticleDOI
01 Oct 1992
TL;DR: This work presents a cascade adaptive filter to remove the baseline wander in the ECG preserving the overlapped deterministic low frequency components of theECG, such as ST segment components.
Abstract: Baseline wander removal is a classical problem in ECG signal processing. We present a cascade adaptive filter to remove the baseline wander in the ECG preserving the overlapped deterministic low frequency components of the ECG, such as ST segment components. This cascade adaptive filter works in two stages. The first stage is an adaptive notch filter at zero frequency. The second stage is an adaptive impulse correlated filter that estimates the ECG signal correlated with the QRS occurrence. In both stages the LMS algorithm is used with different gain constants μ 1 and μ 2 . We analyse the frequency response of the filter as a function of the μ 1 and μ 2 parameters, selecting those more appropriated for baseline removal. Finally, the performance of the filter is studied on an actual ECG affected by baseline drift.

Journal ArticleDOI
TL;DR: It is shown that the performance of the systolic array is similar to that of a conventional LMS implementation for a wide range of practical conditions.
Abstract: A systolic array design for an adaptive filter is presented. The filter is based on the least-mean-square algorithm, but due to the problems in implementation of the systolic array, a modified algorithm, a special case of the delayed LMS (DLMS), is used. The DLMS algorithm introduces a delay in the updating of the filter coefficients. The convergence and steady-state behavior of the systolic array are analyzed. It is shown that the performance of the systolic array is similar to that of a conventional LMS implementation for a wide range of practical conditions. >

Patent
09 Jul 1992
TL;DR: In this article, a main adaptive digital filter and a sub-adaptive digital filter are provided, and these two adaptive digital filters share a filter coefficient to be controlled, on the side of the main adaptive filter, the shared coefficient is updated so that the difference between the output and a desired response is minimized.
Abstract: A main adaptive digital filter and a sub adaptive digital filter are provided, and these two adaptive digital filters share a filter coefficient to be controlled. On the side of the main adaptive digital filter, the shared coefficient is updated so that the difference between the output and a desired response is minimized and on the side of the subadapted digital filter, the above-stated shared filter coefficient is updated so that the output is minimized. A prescribed limitation is given to the frequency characteristic of a filter coefficient to be adapted, by treating the input of the sub adaptive digital filter as a signal weighted on the frequency or a noise having its band limited with respect to the input signal or the output signal of the main adaptive digital filter, and coefficient updating control is conducted so that the coefficient will not go beyond the limitation.

Journal ArticleDOI
TL;DR: An adaptive filtering algorithm is introduced which employs a quasi-Newton approach to give rapid convergence even with colored inputs and appears to be quite robust in finite-precision implementations.
Abstract: The convergence rate of an adaptive system is closely related to its ability to track a time-varying optimum. Basic adaptive filtering algorithms give poor convergence performance when the input to the adaptive system is colored. More sophisticated algorithms which converge very rapidly regardless of the input spectrum algorithms typically require O(N/sup 2/) computation, where N is the order of the adaptive filter, a significant disadvantage for real-time applications. Also, many of these algorithms behave poorly in finite-precision implementation. An adaptive filtering algorithm is introduced which employs a quasi-Newton approach to give rapid convergence even with colored inputs. The algorithm achieves an overall computational requirement of O(N) and appears to be quite robust in finite-precision implementations. >

Journal ArticleDOI
TL;DR: A 2D recursive low-pass filter with adaptive coefficients for restoring images degraded by Gaussian noise is proposed and can easily be extended so that simultaneous noise removal and edge enhancement is possible.
Abstract: A 2D recursive low-pass filter with adaptive coefficients for restoring images degraded by Gaussian noise is proposed. Some of the ideas developed are also submitted for nonGaussian noise. The adaptation is performed with respect to three local image features-edges, spots, and flat regions-for which detectors are developed by extending some existing methods. It is demonstrated that the filter can easily be extended so that simultaneous noise removal and edge enhancement is possible. A comparison with other approaches is made. Some examples illustrate the performance of the filter. >

Patent
Donald R. Hiller1
12 May 1992
TL;DR: In this paper, the adaptive filter coefficients are updated in response to these correlations so as to minimize correlation of the resultant signal with the noise signal, which is achieved by superimposing on the filter input signal a known noise signal.
Abstract: Coefficients of an adaptive filter are updated continuously during the filter's normal operation. This is achieved by superimposing on the filter input signal a known noise signal. At the filter's output, a counterpart of this known noise signal is subtracted and the resultant signal is cross correlated with past samples of the noise signal. The filter coefficients are then updated in response to these correlations so as to minimize correlation of the resultant signal with the noise signal.

01 Jan 1992
TL;DR: In this paper, a new adaptive filter structure is introduced that permits a closer placement of the transducers and that allows the cancellation of noise in the presence of crosstalk.
Abstract: The application of adaptive filters in noise cancel- ing often requires the relative placement of the two transducers at a distance that necessitates a large order filter in order to obtain an adequate output signal-to-noise ratio. A new adaptive filter structure is introduced that permits a closer placement of the transducers and that allows the cancellation of noise in the presence of crosstalk. Algorithms are developed for the new transversal and lattice filter estimators. Simulations show con- siderable improvement in mean-square error over that obtained with standard noise canceling algorithms.

Patent
19 Nov 1992
TL;DR: The disclosed adaptive finite impulse response (FIR) digital filter architecture as mentioned in this paper computes tap coefficient updates in parallel with and simultaneously with the computation of the filter output at each iteration, in a pattern that requires that they be fetched only once from their respective memories.
Abstract: The disclosed adaptive finite impulse response (FIR) digital filter architecture computes tap coefficient updates in parallel with and simultaneously with the computation of the filter output at each iteration. The filter includes a filter output processor (206) and a tap update processor (212) which respectively process, in parallel, the filter output and the coefficient updates for a subsequent iteration. The filter output processor and the tap update processor form their respective outputs at each iteration from input signal values at previous iterations stored in a filter input memory (203) and from tap coefficients stored in a filter taps memory (205). In even-numbered iterations only the even-numbered taps are updated and in odd-numbered iterations only the odd-numbered taps are updated. At every iteration, in forming the filter output and the even or odd tap updates, the previous inputs and tap coefficients are accessed in a pattern that requires that they be fetched only once from their respective memories.

Patent
Akihiko Sugiyama1
20 Feb 1992
TL;DR: In this article, a method or apparatus for controlling coefficients of an adaptive filter (3) for identifying unknown system or predicting periodic signals by correcting coefficients of the adaptive filter(3) in such a manner that the difference signal obtained by subtracting an output signal of the Adaptive Filter from a mixed signal of output signal from the unknown system and an interference signal comprises steps or means (9) for obtaining the information relating to the magnitude of the coefficients or output, and adaptively varying the amount of correction in coefficients of adaptive filter in response to the obtained information.
Abstract: A method or apparatus for controlling coefficients of an adaptive filter (3) for identifying unknown system or predicting periodic signals by correcting coefficients of the adaptive filter (3) in such a manner that the difference signal obtained by subtracting an output signal of the adaptive filter (3) from a mixed signal of the output signal from the unknown system and an interference signal comprises steps or means (9) for obtaining the information relating to the magnitude of the coefficients or output of the adaptive filter, and adaptively varying the amount of correction in coefficients of the adaptive filter in response to the obtained information

Journal ArticleDOI
TL;DR: A conditional median filter that is able to preserve significantly more image detail than the conventional median filter when suppression of impulsive noise is desired by means of bypassing filtering when the local signal variation is below an adaptively adjusted threshold is proposed.
Abstract: We propose a conditional median filter that is able to preserve significantly more image detail than the conventional median filter when suppression of impulsive noise is desired. It does so by means of bypassing filtering when the local signal variation is below an adaptively adjusted threshold. Another advantage of our algorithm is that it is simple to implement and it can be used with any of the available fast median filter algorithms. In addition, it can be easily integrated into existing median filter hardware. We compare the complexity of our algorithm with that of another proposed algorithm with similar filtering objectives. The performance of the new algorithm on real images is compared to that of median and center-weighted median filters.

Journal ArticleDOI
TL;DR: An adaptive notch filter is investigated for eliminating sinusoids imbedded in noise and converges rapidly and attains the Cramer-Rao bound (CRB) for a sufficient large data set.

Proceedings ArticleDOI
10 May 1992
TL;DR: In this article, the authors formulated the filter bank design problem as a quadratic-constrained least-square minimization problem and designed the cosine-modulated and two-channel linear-phase filter banks using the proposed formulation.
Abstract: The author formulates the filter bank design problem as a quadratic-constrained least-square minimization problem. The solution of the minimization problem converges very fast since the cost function and the constraints are quadratic functions with respect to the unknown parameters. The cosine-modulated and the two-channel linear-phase filter banks are designed using the proposed formulation. Compared to other design methods, the proposed technique yields PR (perfect reconstruction) filter banks with much higher stopband attenuation. The filter designed using the new approximation could be used as an initialization filter in a conventional PR filter bank design. >

Proceedings ArticleDOI
10 May 1992
TL;DR: It is shown that the TDOBA clearly outperforms the TDBLMS algorithm with respect to convergence speed and accuracy of adaptation.
Abstract: A technique for 2-D system identification which processes 2-D signals using 2-D blocks is proposed. Two algorithms which perform 2-D FIR (finite impulse response) adaptive filtering using 2-D error blocks or windows are presented. The first algorithm uses a convergence factor that is constant for each 2-D coefficient at each window iteration. This algorithm is termed the two-dimensional block least mean square algorithm (TDBLMS). A novel 2-D adaptive fast LMS algorithm which processes 2-D signals is presented. In this algorithm, a convergence factor is obtained that is the same for all 2-D coefficients at a particular window iteration, but is updated at each window iteration. This algorithm is called the two-dimensional optimum block algorithm (TDOBA). The convergence properties of the TDBLMS and TDOBA are investigated and compared using computer simulations for both disjoint and overlapping windows. It is shown that the TDOBA clearly outperforms the TDBLMS algorithm with respect to convergence speed and accuracy of adaptation. >

Journal ArticleDOI
TL;DR: An adaptive multistage median filter is proposed for preserving the details of images and it is shown that this filter can smooth noise more efficiently than the multistages median filter.
Abstract: An adaptive multistage median filter is proposed for preserving the details of images. It is shown that this filter can smooth noise more efficiently than the multistage median filter. >

Journal ArticleDOI
TL;DR: An efficient approach for the computation of the optimum convergence factor for the LMS (least mean square)/Newton algorithm applied to a transversal FIR structure is proposed, resulting in a dramatic reduction in convergence time.
Abstract: An efficient approach for the computation of the optimum convergence factor for the LMS (least mean square)/Newton algorithm applied to a transversal FIR structure is proposed. The approach leads to a variable step size algorithm that results in a dramatic reduction in convergence time. The algorithm is evaluated in system identification applications where two alternative implementations of the adaptive filter are considered: the conventional transversal FIR realization and adaptive filtering in subbands. >

Journal ArticleDOI
TL;DR: In this article, a multidimensional point-process filter was developed to estimate the actual mode of the target in the presence of the clutter, and the structure and updating properties of the filter have been studied.
Abstract: A discrete-time model of an image-based observation channel using point processes is presented in the context of target-tracking improvement. A multidimensional point-process filter was developed to effectively estimate the actual mode of the target in the presence of the clutter. The structure and updating properties of the filter have been studied. The sample behavior and the robustness, with respect to modeling parameters, are illustrated by simulation examples. >

Proceedings ArticleDOI
10 May 1992
TL;DR: A special class of Nyquist systems is introduced, and an optimum low-pass filter transfer function for the polyphase filter bank of the OMC data transmission system is given.
Abstract: Based on a polyphase IFFT filter bank in the transmitter and a polyphase FFT filter bank in the receiver, orthogonal multiple carrier (OMC) data transmission with high computational efficiency and high bandwidth efficiency can be achieved. The orthogonality is conditioned by an ideal sampling of the receive vector at proper time instants. The authors discuss the effect of the filter transfer function characteristic on the amount of signal distortion due to nonideal timing recovery (timing synchronization) in the receiver. A special class of Nyquist systems is introduced, and an optimum low-pass filter transfer function for the polyphase filter bank of the OMC data transmission system is given. >

Proceedings ArticleDOI
24 Jun 1992
TL;DR: The proposed control scheme together with the optimal Preview filter is shown to be very effective in achieving precision tracking control of discrete time MIMO non-minimum phase systems and the tracking performance is improved as the order of the the preview filter is increased.
Abstract: In this paper, a precision tracking control scheme for linear discrete time non-minimum phase systems is proposed. This control scheme consists of a preview filter, a tracking-performance filter, a command feedforward controller, and a feedback controller. The use of the command feedforward controller, whose design is based on the minimal order inverse model of the plant being controlled, will result in a completely decoupled system. The preview filter is introduced to compensate the phase and gain errors induced by the non-minimum phase zeros or lightly damped zeros of the system. Using the command feedforward controller along with the proposed preview filter, the tracking performance of the proposed control scheme can be characterized by the frequency response of the the tracking-performance filter. For the design of the preview filter, a generalized Nth order preview filter and its associated penalty function that quantifies the tracking error of a design are defined. It is shown that, given the desired bandwidth and the order of the preview filter, the optimal solution for the design of the preview filter can be obtained explicitly. The proposed control scheme together with the optimal preview filter is shown to be very effective in achieving precision tracking control of discrete time MIMO non-minimum phase systems. It is also shown that the tracking performance is improved as the order of the the preview filter is increased.

Journal ArticleDOI
01 Oct 1992
TL;DR: In this paper, a new technique for digital filter design is presented based on the singular value decomposition of the Hankel matrix, balanced realisation and all-pass functions, an IIR filter is obtained via an optimal Hankel-norm approximation.
Abstract: A new technique for digital filter design is presented. Based on the singular value decomposition of the Hankel matrix, balanced realisation and all-pass functions, an IIR filter is obtained via an optimal Hankel–norm approximation. The error between the optimal filter with order r and the desired filter is found to be equal to the (r + l)th singular value of the Hankel matrices. The designed low-pass filter and the differentiator are given to illustrate the proposed design algorithm.

Journal ArticleDOI
TL;DR: A split-path adaptive filter is proposed for extracting the model parameters of an autoregressive process and can provide a much faster rate of convergence at the expense of only a moderate increase in computation.
Abstract: A split-path adaptive filter is proposed for extracting the model parameters of an autoregressive process. The structure is composed of two linear phase filters connected in parallel, one antisymmetric and the other symmetric. The two filters are adapted independently on a sample-by-sample basis using the least-mean-square (LMS) algorithm. The performance of the system in terms of convergence speed and excess mean square error is analyzed in detail, and comparisons with the conventional transversal structure are made. Theoretical analysis and experimental results show that the model can provide a much faster rate of convergence at the expense of only a moderate increase in computation. Two methods for choosing control parameters for the split-path adaptive filter are also suggested to improve further the convergence behavior. >