Kernel adaptive filter

About: Kernel adaptive filter is a research topic. Over the lifetime, 8771 publications have been published within this topic receiving 142711 citations.

Decoding apparatus and decoding method

Abstract: A decoding apparatus having a de-ringing filter to filter image data decoded from encoded image data by orthogonal transformation encoding. In the de-ringing filter, a subtracter generates an absolute value of difference between a value of a filter object pixel and a value of at least one pixel selected from pixels surrounding the filter object pixel on the image data. A comparator compares the absolute value with a threshold. A selector outputs the value of the at least one pixel if the absolute value is less than the threshold, and outputs the value of the filter object pixel if the absolute value is not less than the threshold. A convolution operator convolutes a filter coefficient with the value output from the selector, and outputs a convolution result as a filtered value of the filter object pixel.

The marginalized particle filter in practice

Abstract: The marginalized particle filter is a powerful combination of the particle filter and the Kalman filter, which can be used when the underlying model contains a linear sub-structure, subject to Gaussian noise This paper will illustrate several positioning and target tracking applications, solved using the marginalized particle filter Furthermore, we analyze several properties of practical importance, such as its computational complexity and how to cope with quantization effects

A new method of passive harmonic filter planning for controlling voltage distortion in a power system

Abstract: This paper presents a new method for planning single-tuned passive harmonic filters to control harmonic voltage and voltage distortion throughout a power system. Several alternative objective functions are considered as performance indices in the filter planning problem while the IEEE-519 individual and total harmonic voltage distortion limits at each network bus, as well as filter component limits, are modeled as constraints. The tuned frequency deviation of the filter caused by component manufacturing errors and environmental changes is also taken into account. To solve the problem, a two-step procedure is first proposed to place the filters. Next, the planning problem is formulated as a constrained optimization problem for minimizing the defined network objective function and is then solved by a genetic algorithm-based optimizer to obtain the optimal size of each filter component. The usefulness of the proposed method is tested with an actual distribution network. Results show that the method is effective, computationally robust, and is suitable for the passive filter planning in a power system.

Identification of sparse impulse response systems using an adaptive delay filter

Abstract: This paper discusses a new adaptive technique for system identification of a model with a sparse impulse response. The standard technique for adaptive identification uses a filter with adaptive coefficients that range from a coefficient for the zero delay tap through a coefficient for the largest delay tap that is assumed to be in the system. When the system to be modelled has a sparse impulse response, many of these coefficients will converge to zero, or very small values. Since each coefficient is updated in each iteration, a good deal of the adaptation is spent updating these unnecessary weights. The technique presented in this paper represents a new technique that serially adapts each delay tap value as well as the coefficient value. Thus, using this new technique the number of delay taps can be greatly reduced if the system has a sparse impulse response.

Gaussian Particle Implementations of Probability Hypothesis Density Filters

Abstract: The probability hypothesis density (PHD) filter is a multiple-target filter for recursively estimating the number of targets and their state vectors from sets of observations. The filter is able to operate in environments with false alarms and missed detections. Two distinct algorithmic implementations of this technique have been developed. The first of which, called the Particle PHD filter, requires clustering techniques to provide target state estimates which can lead to inaccurate estimates and is computationally expensive. The second algorithm, called the Gaussian Mixture PHD (GM-PHD) filter does not require clustering algorithms but is restricted to linear-Gaussian target dynamics, since it uses the Kalman filter to estimate the means and covariances of the Gaussians. Extensions for the GM-PHD filter allow for mildly non-linear dynamics using extended and Unscented Kalman filters. A new particle implementation of the PHD filter which does not require clustering to determine target states is presented here. The PHD is approximated by a mixture of Gaussians, as in the GM-PHD filter but the transition density and likelihood function can be non-linear. The resulting filter no longer has a closed form solution so Monte Carlo integration is applied for approximating the prediction and update distributions. This is calculated using a bank of Gaussian particle filters, similar to the procedure used with the Gaussian sum particle filter. The new algorithm is derived here and presented with simulated results.