/*
* File: AutoRegressiveMovingAverageFilter.java
* Authors: Kevin R. Dixon
* Company: Sandia National Laboratories
* Project: Cognitive Foundry
*
* Copyright Feb 24, 2009, Sandia Corporation.
* Under the terms of Contract DE-AC04-94AL85000, there is a non-exclusive
* license for use of this work by or on behalf of the U.S. Government.
* Export of this program may require a license from the United States
* Government. See CopyrightHistory.txt for complete details.
*
*/
package gov.sandia.cognition.math.signals;
import gov.sandia.cognition.annotation.PublicationReference;
import gov.sandia.cognition.annotation.PublicationType;
import gov.sandia.cognition.collection.FiniteCapacityBuffer;
import gov.sandia.cognition.evaluator.AbstractStatefulEvaluator;
import gov.sandia.cognition.math.matrix.Vector;
import gov.sandia.cognition.math.matrix.VectorFactory;
import gov.sandia.cognition.util.DefaultPair;
/**
* A type of filter using a moving-average calculation. That is, a finite
* window of inputs are scaled by a (possibly) unique coefficient and then
* summed together. In other words,
* <BR>
* y(n) + a(0)y(n-1) + a(1)y(n-2) + ... + a(p)y(n-p+1)
* <BR>
* = b(0)x(n) + b(1)x(n-1) + ... + b(m)x(n-m).
* <BR>
* Note that this is slightly different than MATLAB/octave's implementation
* in that the first autoregressive coefficient (a(0)) scales y(n-1) and not
* every value.
*
* @author Kevin R. Dixon
* @since 3.0
*/
@PublicationReference(
author="Wikipedia",
title="Infinite impulse response",
type=PublicationType.WebPage,
year=2009,
url="http://en.wikipedia.org/wiki/Infinite_impulse_response"
)
public class AutoRegressiveMovingAverageFilter
extends AbstractStatefulEvaluator<Double,Double,DefaultPair<FiniteCapacityBuffer<Double>,FiniteCapacityBuffer<Double>>>
implements DiscreteTimeFilter<DefaultPair<FiniteCapacityBuffer<Double>,FiniteCapacityBuffer<Double>>>
{
/**
* Coefficients of the moving-average filter. Element 0 is applied to the
* most-recent input, Element 1 is applied to the second-most-recent,
* and so forth. The dimensionality of the Vector is the order of the
* filter.
*/
private Vector movingAverageCoefficients;
/**
* Coefficients of the autoregressive filter. Element 0 is applied to the
* most-recent output, Element 1 is applied to the second-most-recent,
* and so forth. The dimensionality of the Vector is the order of the
* filter.
*/
private Vector autoRegressiveCoefficients;
/**
* Creates a new instance of AutoRegressiveMovingAverageFilter
* @param numAutoregressiveCoefficients
* Number of autoregressive coefficients.
* @param numMovingAverageCoefficients
* Number of moving-average coefficients.
*/
public AutoRegressiveMovingAverageFilter(
int numAutoregressiveCoefficients,
int numMovingAverageCoefficients )
{
this( VectorFactory.getDefault().createVector(
numAutoregressiveCoefficients, 1.0/numAutoregressiveCoefficients ),
VectorFactory.getDefault().createVector(
numMovingAverageCoefficients, 1.0/numMovingAverageCoefficients ) );
}
/**
* Creates a new instance of AutoRegressiveMovingAverageFilter
* @param autoRegressiveCoefficients
* Coefficients of the autoregressive filter. Element 0 is applied to the
* most-recent output, Element 1 is applied to the second-most-recent,
* and so forth. The dimensionality of the Vector is the order of the
* filter.
* @param movingAverageCoefficients
* Coefficients of the moving-average filter. Element 0 is applied to the
* most-recent input, Element 1 is applied to the second-most-recent,
* and so forth. The dimensionality of the Vector is the order of the
* filter.
*/
public AutoRegressiveMovingAverageFilter(
double[] autoRegressiveCoefficients,
double[] movingAverageCoefficients )
{
this( VectorFactory.getDefault().copyArray( autoRegressiveCoefficients ),
VectorFactory.getDefault().copyArray( movingAverageCoefficients ) );
}
/**
* Creates a new instance of AutoRegressiveMovingAverageFilter
* @param autoRegressiveCoefficients
* Coefficients of the autoregressive filter. Element 0 is applied to the
* most-recent output, Element 1 is applied to the second-most-recent,
* and so forth. The dimensionality of the Vector is the order of the
* filter.
* @param movingAverageCoefficients
* Coefficients of the moving-average filter. Element 0 is applied to the
* most-recent input, Element 1 is applied to the second-most-recent,
* and so forth. The dimensionality of the Vector is the order of the
* filter.
*/
public AutoRegressiveMovingAverageFilter(
Vector autoRegressiveCoefficients,
Vector movingAverageCoefficients )
{
this.setAutoregressiveCoefficients( autoRegressiveCoefficients );
this.setMovingAverageCoefficients( movingAverageCoefficients );
}
public DefaultPair<FiniteCapacityBuffer<Double>, FiniteCapacityBuffer<Double>> createDefaultState()
{
return DefaultPair.create(
new FiniteCapacityBuffer<Double>( this.getNumMovingAverageCoefficients() ),
new FiniteCapacityBuffer<Double>( this.getNumAutoRegressiveCoefficients() ) );
}
public Double evaluate(
Double input )
{
double yn = 0.0;
this.getState().getFirst().addFirst( input );
int n = 0;
for( Double xn : this.getState().getFirst() )
{
final double bn = this.getMovingAverageCoefficients().getElement( n );
yn += bn*xn;
n++;
}
n = 0;
for( Double ynm1 : this.getState().getSecond() )
{
final double an = this.getAutoRegressiveCoefficients().getElement( n );
yn -= an*ynm1;
n++;
}
this.getState().getSecond().addFirst( yn );
return yn;
}
@Override
public AutoRegressiveMovingAverageFilter clone()
{
AutoRegressiveMovingAverageFilter clone =
(AutoRegressiveMovingAverageFilter) super.clone();
clone.setAutoregressiveCoefficients(
this.getAutoRegressiveCoefficients().clone() );
clone.setMovingAverageCoefficients(
this.getMovingAverageCoefficients().clone() );
return clone;
}
public Vector convertToVector()
{
return this.getAutoRegressiveCoefficients().stack(
this.getMovingAverageCoefficients() );
}
public void convertFromVector(
Vector parameters )
{
int M = this.getNumAutoRegressiveCoefficients();
int N = this.getNumMovingAverageCoefficients();
if( (M+N) != parameters.getDimensionality() )
{
throw new IllegalArgumentException(
"Number of dimensions of the parameter Vector aren't equal to the number expected." );
}
this.setAutoregressiveCoefficients( parameters.subVector( 0, M-1 ) );
this.setMovingAverageCoefficients( parameters.subVector( M, N+M-1 ) );
}
/**
* Returns the number of coefficients in the moving-average filter.
* @return
* Number of coefficients in the moving-average filter.
*/
public int getNumMovingAverageCoefficients()
{
return (this.getMovingAverageCoefficients() == null)
? 0 : this.getMovingAverageCoefficients().getDimensionality();
}
/**
* Setter for movingAverageCoefficients
* @return
* Coefficients of the moving-average filter. Element 0 is applied to the
* most-recent input, Element 1 is applied to the second-most-recent,
* and so forth. The dimensionality of the Vector is the order of the
* filter.
*/
public Vector getMovingAverageCoefficients()
{
return this.movingAverageCoefficients;
}
/**
* Setter for movingAverageCoefficients
* @param movingAverageCoefficients
* Coefficients of the moving-average filter. Element 0 is applied to the
* most-recent input, Element 1 is applied to the second-most-recent,
* and so forth. The dimensionality of the Vector is the order of the
* filter.
*/
public void setMovingAverageCoefficients(
Vector movingAverageCoefficients )
{
this.movingAverageCoefficients = movingAverageCoefficients;
}
/**
* Gets the number of autoregressive coefficients.
* @return
* Number of autoregressive coefficients.
*/
public int getNumAutoRegressiveCoefficients()
{
return (this.getAutoRegressiveCoefficients() == null)
? 0 : this.getAutoRegressiveCoefficients().getDimensionality();
}
/**
* Getter for autoregressiveCoefficients
* @return
* Coefficients of the autoregressive filter. Element 0 is applied to the
* most-recent output, Element 1 is applied to the second-most-recent,
* and so forth. The dimensionality of the Vector is the order of the
* filter.
*/
public Vector getAutoRegressiveCoefficients()
{
return this.autoRegressiveCoefficients;
}
/**
* Setter for autoregressiveCoefficients
* @param autoRegressiveCoefficients
* Coefficients of the autoregressive filter. Element 0 is applied to the
* most-recent output, Element 1 is applied to the second-most-recent,
* and so forth. The dimensionality of the Vector is the order of the
* filter.
*/
public void setAutoregressiveCoefficients(
Vector autoRegressiveCoefficients )
{
this.autoRegressiveCoefficients = autoRegressiveCoefficients;
}
}