An initial-value solution of the least-squares estimation problem with degenerate covariance (Q792286)

Considered is the following Wiener-Hopf type Fredholm integral equation \[ h(t,s)=k(t,s)-\int^{t}_{0}k(y,s)h(t,y)dy \] for the linear least- square estimation of z(t) in the stochastic process \(y(t)=z(t)+v(t)\). It is found that via a Bellman-Krein-Sobolev type formula the h-function can be expressed in terms of the auxiliary function whose initial-value solution consists of the reflection function. It satisfies a Riccati type nonlinear differential equation. It is shown that a real-time solution of the linear least-square filtering problem is provided by the Cauchy system of the time-integral of the auxiliary function multiplied by the observed signal.