/ 
DSP-LinPred-0.06
River stage zero No dependents
/ DSP::LinPred

NAME

DSP::LinPred - Linear Prediction

SYNOPSIS

use DSP::LinPred;

# OPTIONS
# mu       : Step size of filter. (default = 0.001)
#
# h_length : Filter size. (default = 100)
# dc_mode  : Direct Current Component estimation.
#            it challenges to estimating DC component when set 1.
#            (default = 1 enable)
# dc_init  : Initial DC bias.
#            It *SHOULD* be set value *ACCURATELY* when dc_mode => 0.
#            (default = 0)
#
# stddev_mode : Step size correction by stddev of input.
#               (default = 1 enable)
# stddev_init : Initial value of stddev.
#               (default = 1)
#

my $lp = DSP::LinPred->new;

# set filter
$lp->set_filter({
                 mu => 0.001,
                 filter_length => 500,
                 dc_mode => 1,
                 stddev_mode => 1
                });

# defining signal x
my $x = [0,0.1,0.5, ... ]; # input signal

# Updating Filter
$lp->update($x);
my $current_error = $lp->current_error; # get error

# Prediction
my $pred_length = 10;
my $pred = $lp->predict($pred_length);
for( 0 .. $#$pred ){ print $pred->[$_], "\n"; }

DESCRIPTION

DSP::LinPred is Linear Prediction by Least Mean Squared Algorithm.

This Linear Predicting method can estimate the standard deviation, direct current component, and predict future value of input.

METHODS

set_filter

set_filter method sets filter specifications to DSP::LinPred object.

$lp->set_filter(
    {
        mu => $step_size, # <Num>
        filter_length => $filter_length, # <Int>
        dc_init => $initial_dc_bias, # <Num>
        dc_mode => $dc_estimation, # <Int>, enable when 1
        stddev_init => $initial_stddev, # <Num>
        stddev_mode => $stddev_estimation # <Int>, enable when 1
    });

update

update method updates filter state by source inputs are typed ArrayRef[Num].

my $x = [0.13,0.3,-0.2,0.5,-0.07];
$lp->update($x);

If you would like to extract the filter state, you can access member variable directly like below.

my $filter = $lp->h;
for( 0 .. $#$filter ){ print $filter->[$_], "\n"; }

predict

predict method generates predicted future values of inputs by filter.

my $predicted = $lp->predict(7);
for( 0 .. $#$predicted ){ print $predicted->[$_], "\n";}

filter_dc

This method can calculate mean value of current filter.

my $filter_dc = $lp->filter_dc;

filter_stddev

This method can calculate standard deviation of current filter.

my $filter_stddev = $lp->filter_stddev;

READING STATES

current_error

# It returns value of current prediction error
# error = Actual - Predicted
my $current_error = $lp->current_error;
print 'Current Error : '.$current_error, "\n";

h

# It returns filter state(ArrayRef)
my $filter = $lp->h;
print "Filter state\n";
for( 0 .. $#$filter ){ print $_.' : '.$filter->[$_],"\n"; }

x_count

# It returns value of input counter used in filter updating.
my $x_count = $lp->x_count;
print 'Input count : '.$x_count, "\n";

dc

# Get value of current Direct Current Components of inputs.
my $dc = $lp->dc;
print 'Current DC-Component : '.$dc, "\n";

stddev

# Get value of current standard deviation of inputs.
my $stddev = $lp->dc;
print 'Current STDDEV : '.$stddev, "\n";

EXPERIMENTAL OPTIONS

iir_mode and iir_a

In set_filter option. if set iir_mode to 1, and iir_a, it challenges to calculate DC value and stddev by using IIR filter.

IIR spec: next_dc = (current_input - iir_a * current_dc) / (1 - iir_a)

next_stddev = (abs(current_input - current_dc) - iir_a * current_stddev) / (1 - iir_a)

# set_filter with iir_mode on.
$lp->set_filter({
                 mu => 0.001,
                 filter_length => 500,
                 dc_mode => 1,
                 stddev_mode => 1,
                 iir_mode => 1,
                 iir_a => 0.01
                });

LICENSE

Copyright (C) Toshiaki Yokoda.

This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself.

AUTHOR

Toshiaki Yokoda <>