Ladder-form filters for nonstationary processes

Ladder-form whitening filters for discrete-time non-stationary processes are presented. Two fundamental versions, both cascade connections of identical stages are described. One version has a simple filter stage, but time varying gains which require a separate, parallel computation; the other has a more complex filter stage, but time invariant parameters. The overall complexity of both schemes is proportional to the index of nonstationarity of the underlying stochastic process. The resulting recursive relations are applicable to problems of covariance factorization, spectral estimation and parameter identification.