We investigate multiplicative noise models of the form z(t) = s(t)g(t)
+ n(t), where z(t) is the observed process. In this model, s(t), g(t)
and n(t) are mutually independent stationary random processes, s(t) i
s the signal process, g(t) is the multiplicative noise process, and n(
t) is a colored Gaussian process. The objective is to estimate some st
atistical properties of the signal process, s(t), from observations of
z(t). If s(t) depends upon some non-random parameter vector, theta, t
hen an additional objective is to estimate the vector theta. We develo
p algorithms to estimate the signal parameters, when s(t) is a linear
process or a harmonic process; these algorithms are based on the highe
r-order moments and cumulants of the observed process.