Filter Stability for Stochastic Evolution Equations

It is established that the Kalman filter associated with signal and observation processes defined through stochastic evolution equations is stable under very weak hypotheses; namely when appropriate stabilizability / detectability criteria hold. Thus in this general setting we obtain results as sharp as are available for processes taking values in finite dimensional linear spaces. The conditions are shown to be directly verifiable in certain important situations.