scispace - formally typeset
Search or ask a question

Showing papers on "Kernel adaptive filter published in 1988"


Proceedings ArticleDOI
C.W. Farrow1
07 Jun 1988
TL;DR: An FIR (finite-impulse-response) filter which synthesizes a controllable delay which has the ability to interpolate between samples in the data stream of a band-limited signal is described.
Abstract: The author describes an FIR (finite-impulse-response) filter which synthesizes a controllable delay. By changing the delay the filter has the ability to interpolate between samples in the data stream of a band-limited signal. Because high sampling rates are not required, the filter is especially suited for implementation in a digital signal processor (DSP), and has been implemented in a real-time DSP. The interpolator can be used as a practical way to reconstruct an original band limited signal from samples taken at the Nyquist rate. The variable delay filter can also be used as a more general computational element. Performance results are presented. >

853 citations


Journal ArticleDOI
TL;DR: An adaptive smoothing filter is proposed for reducing noise in digital signals of any dimensionality based on the selection of an appropriate inner or outer trimmed mean filter according to local measurements of the tail behavior (impulsivity) of the noise process.
Abstract: An adaptive smoothing filter is proposed for reducing noise in digital signals of any dimensionality. The adaptive procedure is based on the selection of an appropriate inner or outer trimmed mean filter according to local measurements of the tail behavior (impulsivity) of the noise process. The set of trimmed means used provides robustness against a wide range of noise possibilities ranging from very shallow tailed to very heavy tailed. A Monte Carlo analysis using a family of generalized exponential distributions supports the choice of the trimmed mean selected for measured values of an impulsivity statistic. The assumption underlying the definition of the filter is that the signal to be filtered is locally smoothly varying, and that the noise process is uncorrelated and derives from an unknown, unimodal symmetric distribution. For image-processing applications, a second statistic is used to mark the location of abrupt intensity changes, or edges; in the vicinity of an edge, the trend-preserving median filter is used. Since the impulsivity and edge statistics used in defining the adaptive filter are both functions of order statistics, the extra computation required for their calculation is minimal. Examples are provided of the filter as applied to images corrupted by a variety of noises. >

134 citations


Journal ArticleDOI
TL;DR: It is shown that the alpha -TM and STM filters perform better than the running mean and median filters in white noise suppression, while they can be designed to be comparable to the median filter in edge preservation in the presence of noise.
Abstract: The statistical properties of two classes of filters generalizing the median filter are considered. The two classes are the alpha-trimmed mean ( alpha -TM) filter and the standard type M(STM) filter, both of which are special cases of the L filter. Results are developed to quantify the white-noise suppression and edge-preservation characteristics of the filters by considering their output sequence error statistics. It is shown that the alpha -TM and STM filters perform better than the running mean and median filters in white noise suppression, while they can be designed to be comparable to the median filter in edge preservation in the presence of noise. >

65 citations


Journal ArticleDOI
TL;DR: An adaptive notch filter based on a fast recursive-least-squares (FLS) algorithm is introduced, which has about the same complexity as the fast transversal adaptive filter but performs better for the analysis of narrowband signals in high-level noise.
Abstract: An adaptive notch filter based on a fast recursive-least-squares (FLS) algorithm is introduced. The structure is canonical and consists of a transversal section cascaded with an IIR section. The adaptive procedure takes place in the transversal section, and the impact of the notch factor is pointed out. Overall, the FLS notch filter has about the same complexity as the fast transversal adaptive filter, but it performs better for the analysis of narrowband signals in high-level noise. >

61 citations


Journal ArticleDOI
TL;DR: It is shown that the mean behavior of the adaptive filter coefficients can be studied by using an ordinary differential equation (ODE) and approximate and simple closed form results are derived for the tracking behavior of a second order notch filter.
Abstract: The tracking behavior of a constrained IIR notch filter, the coefficients of which are estimated using a constant-step-size Gauss-Newton algorithm, is studied. Using weak convergence theory and the concept of prescaling, it is shown that the mean behavior of the adaptive filter coefficients can be studied by using an ordinary differential equation (ODE). Approximate and simple closed-form results are derived for the tracking behavior of a second-order notch filter. Computer simulations are presented to substantiate the analysis. >

60 citations


Journal ArticleDOI
01 Jun 1988
TL;DR: This work presents a technique that permits constant-time filtering for space-variant kernels, and allows the use of arbitrary filters, and is useful to explore interesting mappings and special filtering techniques.
Abstract: Filtering is an essential but costly step in many computer graphics applications, most notably in texture mapping. Several techniques have been previously developed which allow prefiltering of a texture (or in general an image) in time that is independent of the number of texture elements under the filter kernel. These are limited, however, to space-invariant kernels whose shape in texture space is the same independently of their positions, and usually are also limited to a small range of filters.We present here a technique that permits constant-time filtering for space-variant kernels. The essential step is to approximate a filter surface in texture space by a sum of suitably-chosen basis functions. The convolution of a filter with a texture is replaced by the weighted sum of the convolution of the basis functions with the texture, which can be precomputed. To achieve constant time, convolutions with the basis functions are computed and stored in a pyramidal fashion, and the right level of the pyramid is selected so that only a constant number of points on the filter kernel need be evaluated.The technique allows the use of arbitrary filters, and as such is useful to explore interesting mappings and special filtering techniques. We give examples of applications to perspective and conformal mappings, and to the use of filters such as gaussians and sinc functions.

51 citations


Journal ArticleDOI
TL;DR: Local stability of the arithmetic version of the constant modulus algorithm of adaptive filtering is proven in two applications: channel equalization for a transmitted sequence of plus and minus ones, and the separation of a sinusoidal signal from its sum with a number ofsinusoidal interferers at separate frequencies.
Abstract: The constant modulus algorithm (CMA) of adaptive filtering was developed for usage when the modulus of the desired signal is known, but its specific value at every sample instant is not known. A real-arithmetic version of CMA was recently proposed, and studied using simulations, by J.R. Treichler and M.G. Larimore (ibid., vol.ASSP-33, p.420-31, Apr. 1985). Local stability of their arithmetic version of CMA is proven in two applications: channel equalization for a transmitted sequence of plus and minus ones, and the separation of a sinusoidal signal from its sum with a number of sinusoidal interferers at separate frequencies. The proofs utilize dynamic system stability theorems from averaging theory, which is a technique currently being utilized in the stability analysis of a variety of adaptive systems. >

44 citations


Proceedings ArticleDOI
11 Apr 1988
TL;DR: A fast, recursive least-squares (RLS) adaptive nonlinear filter is presented that makes use of the ideas of fast RLS multichannel filters and has a computational complexity of O(N/sup 3/) multiplications.
Abstract: A fast, recursive least-squares (RLS) adaptive nonlinear filter is presented. The nonlinearity is modeled using a second-order Volterra-series expansion. The structure makes use of the ideas of fast RLS multichannel filters and has a computational complexity of O(N/sup 3/) multiplications. This compares with O(N/sup 6/) multiplications required for direct implementation. Simulation examples in which the filter is employed to identify nonlinear systems using noisy output observations are also presented. Further simplification to the structure through a simplified model is discussed. >

39 citations


Journal ArticleDOI
TL;DR: The time (shift) delay parameter between two signals is modeled as a finite-impulse response filter whose coefficients are samples of a sinc function, which involves less computation and the elimination of interpolation needed in previous approaches to obtain nonintegral time-delay estimates.
Abstract: The time (shift) delay parameter between two signals is modeled as a finite-impulse response filter whose coefficients are samples of a sinc function. The time-domain LMS (least-mean-squares) adaptive algorithm is used, but only the weight with the largest magnitude is updated, which involves less computation. The result is a faster adaptation and the elimination of interpolation needed in previous approaches to obtain nonintegral (multiples of sampling period) time-delay estimates. >

33 citations


Journal ArticleDOI
TL;DR: The Kalman filter design has been optimized to minimize the program memory requirements and execution speed, and higher system bandwidth can be accommodated through higher-speed digital signal processors.
Abstract: A Kalman filter for tracking moving objects has been implemented on a TMS32010 digital signal processor. Tracking accuracy and quantization effects of the implementation have been measured by comparing the filter to one implemented on a general-purpose computer with a 32 bit word length. The filter design has been optimized to minimize the program memory requirements and execution speed. Although the filter has been implemented on a specific signal processing chip, the design is general enough to be applicable to any other digital signal processor. The filter can be used for tracking objects for industrial or other applications where range and bearing measurements are available. For motion on a plane, the filter can be used to track objects where the maximum system bandwidth is 1680 Hz; for three-dimensional motion the system bandwidth is 1120 Hz. Using the approach presented, higher system bandwidth can be accommodated through higher-speed digital signal processors. >

29 citations


Patent
29 Sep 1988
TL;DR: In this paper, a double-talker detector is used in conjunction with the echo-canceller to detect near-end speech or signal, which is detected if the latter signal exceeds a predetermined function of the former signal.
Abstract: Method and apparatus are disclosed which provide echo-cancellation in subscriber line audio-processing circuits (SLACs) and modulator-demodulators (modems). The echo-canceller provides a response independent of the amplitude of an error signal representing the difference between the desired and generated signal. The convergence of the echo-cancelling is not influenced by the size of the error signal but only by its sign. In a preferred embodiment, a cannonic signed digit filter coefficient updating technique is used for a digital filter (10) implementing the echo-cancellation function. A double-talker detector is advantageously used in conjunction with the echo-canceller. Detection of "double-talker" near-end signals inhibits updating of the echo-canceller filter coefficients. Near-end signals are detected by an energy-averaging filter (26) which selectively samples (28) low-pass filtered signals received by the adaptive filter as received from the far-end talker and low-pass filtered signals received from the near-end talker. Near-end speech or signal is detected if the latter signal exceeds a predetermined function of the former signal.

Proceedings ArticleDOI
07 Jun 1988
TL;DR: In this paper, an adaptive 2D digital filter is presented in which the filter is determined by a 1-D FIR (finite-impulse-response) prototype filter, and hence the computational complexity of the coefficient-update algorithm has an order of complexity that is similar to a 1D adaptive filter.
Abstract: An adaptive 2-D digital filter is presented in which the filter is determined by a 1-D FIR (finite-impulse-response) prototype filter, and hence the computational complexity of the coefficient-update algorithm has an order of complexity that is similar to a 1-D adaptive filter. The convergence rate of the proposed structure is much faster than that of a direct-form 2-D LMS (least-mean-square) filter. Results of the hardware realization of the filter are presented that show its learning curve to be considerably less noisy than that of the 2-D direct-form LMS filter. >

Patent
Tuan H. Bui1
29 Apr 1988
TL;DR: In this paper, an adaptive digital filter can be implemented on a single Very Large Scale Integrated (VLSI) circuit silicon chip and the Least Mean Square adaptive filter algorithm can be performed by parallel processing during a single clock cycle.
Abstract: An adaptive digital filter can be implemented on a single Very Large Scale Integrated (VLSI) circuit silicon chip and the Least Mean Square adaptive filter algorithm can be performed by parallel processing during a single clock cycle. The adaptive filter contains dual delay lines to yield a sequence of simultaneous samples of both input and output signals. Correlations of the present error difference with previous samples of both input and output signals can then take place simultaneously in each clock cycle. The adaptive filter is modular and can be cascaded with other identical filters to form a high-order filter.

Journal ArticleDOI
A.S. Abutaleb1
TL;DR: In this article, a nonlinear filter for adaptive noise cancellation is proposed, which is based on the Pontryagin minimum principle and the method of invariant imbedding, and its computational time is about 10% that of the LMS in the cases studied.
Abstract: The author introduces a nonlinear filter for adaptive noise cancelling. The derivation and convergence properties of the filter are presented. The performance, as measured by the signal-to-noise ratio between the signal and its estimate, is compared to that of the commonly used least-mean-square (LMS) algorithm. It is shown, through simulation, that the proposed canceller has, on the average, better performance than the LMS canceller. It is based on the Pontryagin minimum principle and the method of invariant imbedding, and its computational time is about 10% that of the LMS in the cases studied, which is a substantial improvement. >

Journal ArticleDOI
TL;DR: In this paper, a modified Wiener filter is proposed to suppress additive noise by using statistical information obtained during wavelet estimation, and a value of p can be chosen which gives a modified WF equivalent to a statistically robust deconvolution filter.
Abstract: Deconvolution in the presence of additive noise is a well known problem for which there exists a Wiener filter which spectrally whitens while also suppressing the noise. A simple variant of this standard Wiener filter incorporates a parameter p which is intended to allow further weight to be given to noise suppression. This filter is often called a modified Wiener filter. To design such a filter, one must know the frequency characteristics of the wavelet precisely, plus the spectra of the input and additive noise. Typically, some appropriate estimate of the frequency function of the wavelet is taken, and the modified Wiener filter is designed from that estimate. A more realistic practical viewpoint is to think of the estimated wavelet response as one of a set of possible frequency response functions. By using statistical information obtained during wavelet estimation, a value of p can be chosen which gives a modified Wiener filter equivalent to a statistically robust deconvolution filter. Here “robust” me...

Journal ArticleDOI
TL;DR: In this article, the problem of finite-time, reduced-order, minimum variance full-state estimation of linear, continuous time-invariant systems is considered in cases where the output measurement is partially free of corrupting white-noise components.
Abstract: The problem of the finite-time, reduced-order, minimum variance full-state estimation of linear, continuous time-invariant systems is considered in cases where the output measurement is partially free of corrupting white-noise components. The structure of the optimal filter is obtained and a link between this structure and the structure of the system invariant zeros is established. Using expressions that are derived in closed form for the invariant zeros of the system, simple sufficient conditions are obtained for the existence of the optimal filter in the stationary case. The structure and the transmission properties of the stationary filter for general left-invertible systems are investigated. A direct relation between the optimal filter and a particular minimum-order left inverse of the system is obtained. A simple explicit expression for the filter transfer function matrix is also derived. The expression provides an insight into the mechanism of the optimal estimation. >

Proceedings ArticleDOI
11 Apr 1988
TL;DR: A structure for the subband acoustic echo canceler is proposed that avoids filter banks in the sending line and the consequent delay and an additional adaptive filter is included.
Abstract: A structure for the subband acoustic echo canceler is proposed that avoids filter banks in the sending line and the consequent delay. In this arrangement of the filter banks, the conventional least-mean-square algorithm is modified and an additional adaptive filter is included. Simulation results are given, and a real application is reported. >

Patent
20 Jul 1988
TL;DR: In this paper, a transversal filter has a set of tap gains for producing an equalized signal as a filter output from the digital signal incoming to the transmission line, and a controller has a plurality of sets of values for the set of gains, compares the mean value with a reference value to produce an error signal and determines one set of the plurality of set of values in response to the error signal to make the tap gains equal to the set values thereby to maintain the average value to the reference value.
Abstract: A digital automatic line equalizer for compensating distortion of a digital signal transmitted through a transmission line comprises a transversal filter having a set of tap gains for producing an equalized signal as a filter output from the digital signal incoming thereto. A first multiplier squares a current one of the filter output to produce a squared signal and other multipliers multiply the current and the previous ones of the filter outputs to produce product signals. In a calculator, the product signals are weighted and are, thereafter, summed together with the squared signal to produce summed signals which are averaged for a predetermined time duration into a mean value. A controller has a plurality of sets of values for the set of tap gains, compares the mean value with a reference value to produce an error signal and determines one set of the plurality of sets of values in response to the error signal to make the set of tap gains equal to the set of values thereby to maintain the mean value to the reference value so that the transversal filter produces the equalized signal.

Journal ArticleDOI
TL;DR: A rigorous proof of almost-sure convergence to the optimal filter is attained under a weak-ergodicity assumption that includes the case, important in practice, of correlated observations.
Abstract: The convergence analysis of a sign algorithm with decreasing gain, when governing the weights of an adaptive filter, is presented. The analysis is done in a noiseless case. A rigorous proof of almost-sure convergence to the optimal filter is attained under a weak-ergodicity assumption that includes the case, important in practice, of correlated observations. >

Proceedings ArticleDOI
11 Apr 1988
TL;DR: A class of M-channel FIR (finite-impulse response) perfect-reconstruction filter banks is introduced that includes the recently developed lossless filter banks and the results are simple formulas for required storage and computation, which helps in the selection of efficient tree structures.
Abstract: The concept of perfect reconstruction in filter banks is examined using the Smith form of the polyphase matrix. A class of M-channel FIR (finite-impulse response) perfect-reconstruction filter banks is introduced that includes the recently developed lossless filter banks. With the proposed filter banks, the synthesis filters are of the same complexity as the analysis filters. In addition, the synthesis filter bank is easily obtained from the analysis filter bank by inverting a set of M*M constant coefficient matrices. A statistical method for the design of such filter banks is presented. These can be used as building blocks in large tree-structured filter banks where the overall number of channels is any composite integer. An analysis of the computational and storage complexity of such tree structures is given. The results of the analysis are simple formulas for required storage and computation, which helps in the selection of efficient tree structures. >

Journal ArticleDOI
TL;DR: It is shown that the mean-square deviation is bounded by a constant multiple of the adaptation step size and that the same holds for the excess error of the signal estimation.
Abstract: The convergence properties of an adaptive linear mean-square estimator that uses a modified LMS algorithm are established for generally dependent processes. Bounds on the mean-square error of the estimates of the filter coefficients and on the excess error of the estimate of the signal are derived for input processes which are either strong mixing or asymptotically uncorrelated. It is shown that the mean-square deviation is bounded by a constant multiple of the adaptation step size and that the same holds for the excess error of the signal estimation. The present findings extend earlier results in the literature obtained for independent and M-dependent input data. >

Proceedings ArticleDOI
05 Jun 1988
TL;DR: A PD (partial-discharger) measuring device to improve on-site measurements has been developed that is based on a digital filtering algorithm that suppresses external sinusoidal disturbances effectively.
Abstract: A PD (partial-discharger) measuring device to improve on-site measurements has been developed that is based on a digital filtering algorithm. The adaptive filter reduces periodic interference, e.g. from broadcasting stations. The filter principle is based on a fast Fourier transform that it suppresses external sinusoidal disturbances effectively. Even partial discharges with an apparent charge less than the basic interference level can easily be detected. >

Journal ArticleDOI
TL;DR: In this paper, a one-dimensional Kalman filter algorithm is presented that resolves several overlapped liquid chromatographic peaks without algebraic operations of matrices, and the reliability is shown to be similar to that of the multi-dimensional filter for the resolution of overlapped Gaussian peaks with limited R s and S/N values.

Proceedings ArticleDOI
23 Oct 1988
TL;DR: In this article, a type of adaptive filter called a time-dependent adaptive filter (TDAF) was proposed to deal with the cyclostationary nature of communication signals by changing periodically.
Abstract: The authors examine a type of adaptive filter called a time-dependent adaptive filter (TDAF). A TDAF can effectively deal with the cyclostationary nature of communication signals by changing periodically. It achieves a smaller mean-square error than conventional filter for cyclostationary signals. A frequency-domain TDAF is investigated, and it is shown that it is especially effective for interference rejection applications. Simulation results are presented, showing how cochannel QPSK signals can be separated. >

Journal ArticleDOI
TL;DR: Under rather weak conditions, the algorithm for adaptive filters with contraints or adaptive filters of beam former type is considered and the strong consistency result is obtained.

Proceedings ArticleDOI
07 Jun 1988
TL;DR: A modular VLSI architecture for a novel linear-phase FIR (finite impulse response) filter structure based on an IIR subfilter whose infinite impulse response is truncated into a finite one.
Abstract: The authors introduce a modular VLSI architecture for a novel linear-phase FIR (finite impulse response) filter structure. The proposed linear-phase FIR filters have a highly reduced number of general multiplications per sample compared to conventional FIR filters. The novel filter structure is based on an IIR subfilter whose infinite impulse response is truncated into a finite one. Although the subfilter has a nonlinear phase response, it can be made exactly linear by reversing the data stream in time and using the same filter again. Another possibility is to use a maximum-phase version of the FIR filter in cascade. The choice between different realizations depends on the filter specifications. In the general case the filter coefficients cannot be represented with the simple shift-and-add procedure and the time reversal technique should be used. The authors prefer the maximum-phase FIR alternative. >

Journal ArticleDOI
01 May 1988
TL;DR: The authors update delta modulation digital filter weights using the LMS (least-mean-squares) and the SIGN algorithms to realize an adaptive digital filter without multiplication operations.
Abstract: The authors update delta modulation digital filter weights using the LMS (least-mean-squares) and the SIGN algorithms to realize an adaptive digital filter without multiplication operations. It is shown that using the SIGN algorithm results in an adaptive filter that can be implemented using the simple up/down counting operations. Learning curves demonstrating convergence properties of the algorithms for a system identification problem are presented. >

Proceedings ArticleDOI
07 Jun 1988
TL;DR: In this article, an LMS (least-mean-squares) algorithm is proposed for adapting the poles and zeros of state-space IIR (infinite-impulse-response) filters.
Abstract: The authors propose an LMS (least-mean-squares) algorithm for adapting the poles and zeros of state-space IIR (infinite-impulse-response) filters. Recently presented state-space sensitivity formulae are used to obtain gradients which are required to adapt the filter coefficients. The number of computations for a general state-space adaptive filter is seen to be large, but a modified companion-form filter is shown to require much less computation. Also, it is shown that, in general, it is possible to efficiently adapt any state-space system by modifying any one column of its A matrix. This could prove useful where only small changes in coefficients are expected. Finally, simulation results are presented. >


Journal ArticleDOI
TL;DR: A covariance-RLS lattice adaptive-filtering algorithm that permits nonzero initial conditions and need not 'warn' N-1 iterations in advance of the first iteration to preserve its low computational requirements is presented.
Abstract: A covariance-RLS lattice adaptive-filtering algorithm is presented. The algorithm permits nonzero initial conditions and need not 'warn' N-1 iterations in advance of the first iteration to preserve its low computational requirements. These are both improvements over previous lattice algorithms. The algorithm also has a slight computational advantage over previous solutions, rendering it more applicable to adaptive-filtering applications such as fast-starting adaptive equalizers and echo cancellers, where the initial data in the adaptive filter are not, and cannot be assumed to be, zero. >