----------------------------------------------------------------------------- -- | -- Module : DSP.Filter.IIR.Prony -- Copyright : (c) Matthew Donadio 2003 -- License : GPL -- -- Maintainer : m.p.donadio@ieee.org -- Stability : experimental -- Portability : portable -- -- General case of Prony's Method where K > p+q -- -- References: L&I, Sect 8.1; P&B, Sect 7.5; P&M, Sect 8.5.2 -- -- Notation follows L&I -- ----------------------------------------------------------------------------- -- TODO: Handle rank deficiencies of G3 gracefully. Can/should we -- generate a (K/2+1) by (K/2+1) G2, and set p=q=rank(G2)? Need SVD to -- compute rank, though. module DSP.Filter.IIR.Prony (prony) where import Data.Array import Matrix.Matrix import Matrix.LU {------------------------------------------------------------------------------ Case 1: K=p+q a = array (0,p) b = array (0,q) g1 : q+1 by p+1 g2 : p by p+1 g3 : p by p We do not define G1 and G2, but mg2 = array ((1,1),(p,p+1)) [ ((i,j), g!(p+i+1-j)) | j <- [1..p+1], i <- [1..p] ] prony p q g = (a,b) where mg3 = array ((1,1),(p,p)) [ ((i,j), g!(p+i-j)) | j <- [1..p], i <- [1..p] ] g1 = array (1,p) [ (i, g!(p+i)) | i <- [1..p] ] a' = solve mg3 (fmap negate g1) a = array (0,p) $ (0,1) : [ (i,a'!i) | i <- [1..p] ] b = listArray (0,q) [ sum [ a!j * g!(i-j) | j <- [0..(min i p)] ] | i <- [0..q] ] Test case, pg 422 g = listArray (0,6) [ 1, 18, 9, 2, 1, 2/9, 1/9 ] :: Array Int Double ------------------------------------------------------------------------------} -- Case 2: K>p+q -- a = array (0,p) -- b = array (0,q) -- g1 : q+1 by p+1 -- g2 : K-q by p+1 -- g3 : K-q by p -- We need gi for the q

Int -- ^ q -> Array Int Double -- ^ g[n] -> (Array Int Double, Array Int Double) -- ^ (b,a) prony p q g = (b,a) where k = snd $ bounds g gi i | i < 0 = 0 | i > k = 0 | otherwise = g!i mg3 = array ((1,1),(k-q,p)) [ ((i,j), gi (q+i-j)) | j <- [1..p], i <- [1..k-q] ] g1 = array (1,k-q) [ (i, gi (q+i)) | i <- [1..k-q] ] a' = solve (mm_mult (m_trans mg3) mg3) (fmap negate (mv_mult (m_trans mg3) g1)) a = array (0,p) $ (0,1) : [ (i,a'!i) | i <- [1..p] ] b = listArray (0,q) [ sum [ a!j * gi (i-j) | j <- [0..(min i p)] ] | i <- [0..q] ]