We develop several methods to transform a non-positive real transfer matrix
into a positive real one. The problem is of practical engineering interest
, since it might arise when trying to identify a linear description of a sy
stem, by means of stochastic subspace identification procedures. The modifi
cations proposed preserve rationality and are reasonable in terms of system
s theoretic properties expected of spectral density matrices. First, a stab
ility problem related to stationarity of an underlying stochastic process i
s addressed and solved, making use of the reciprocal symmetry of spectral d
ensities or alternatively Glover's optimal approximation. Then three method
s are presented which compensate possible remaining positivity problems. Th
e first two make use of the Kalman-Yakubovich-Popov lemma and the recent ad
vances in semidefinite programming problems. The last method suggests corre
ctions based on the asymptotic behaviour of generalized Schur parameters an
d algorithms related to maximum entropy extension and the backward Levinson
algorithm. (C) 2000 Elsevier Science B.V. All rights reserved.