Scilab Website | Contribute with GitLab | Mailing list archives | ATOMS toolboxes
Scilab Online Help
5.4.1 - Русский

Change language to:
English - Français - 日本語 - Português -

Please note that the recommended version of Scilab is 2024.0.0. This page might be outdated.
See the recommended documentation of this function

Scilab help >> Signal Processing > identification > phc

phc

Markovian representation

Calling Sequence

[H,F,G]=phc(hk,d,r)

Arguments

hk

hankel matrix

d

dimension of the observation

r

desired dimension of the state vector for the approximated model

H, F, G

relevant matrices of the Markovian model

Description

Function which computes the matrices H, F, G of a Markovian representation by the principal hankel component approximation method, from the hankel matrix built from the covariance sequence of a stochastic process.

Examples

//This example may usefully be compared with the results from 
//the 'levin' macro (see the corresponding help and example)
//
//We consider the process defined by two sinusoids (1Hz and 2 Hz) 
//in additive Gaussian noise (this is the observation); 
//the simulated process is sampled at 10 Hz.

t=0:.1:100;rand('normal');
y=sin(2*%pi*t)+sin(2*%pi*2*t);y=y+rand(y);plot(t,y)

//covariance of y

nlag=128;
c=corr(y,nlag);

//hankel matrix from the covariance sequence
//(we can choose to take more information from covariance
//by taking greater n and m; try it to compare the results !

n=20;m=20;
h=hank(n,m,c);

//compute the Markov representation (mh,mf,mg)
//We just take here a state dimension equal to 4 :
//this is the rather difficult problem of estimating the order !
//Try varying ns ! 
//(the observation dimension is here equal to one)

ns=4;
[mh,mf,mg]=phc(h,1,ns);

//verify that the spectrum of mf contains the 
//frequency spectrum of the observed process y
//(remember that y is sampled -in our example 
//at 10Hz (T=0.1s) so that we need 
//to retrieve the original frequencies through the log 
//and correct scaling by the frequency sampling)

s=spec(mf);s=log(s);
s=s/2/%pi/.1;

//now we get the estimated spectrum
imag(s),

See Also

  • levin — Toeplitz system solver by Levinson algorithm (multidimensional)
Report an issue
<< lattp identification rpem >>

Copyright (c) 2022-2023 (Dassault Systèmes)
Copyright (c) 2017-2022 (ESI Group)
Copyright (c) 2011-2017 (Scilab Enterprises)
Copyright (c) 1989-2012 (INRIA)
Copyright (c) 1989-2007 (ENPC)
with contributors
Last updated:
Tue Apr 02 17:37:55 CEST 2013