Constrained Nonstationary Signal Processing by Pair-wise Separable Quadratic Programming

V. Mottl, O. Krasotkina (Russia), and I. Muchnik (USA)


Multidimensional signal processing, estimation of signal parameters, nonstationary regression, pair-wise separable quadratic programming, dynamic programming


There exists a wide class of nonstationary signal processing problems which invite for formulating them as those of estimating the succession of nonstationary regression parameters constrained at each time moment by linear inequalitites. Such a formalization leads to the so-called pair-wise separable quadratic programming problem whose specificity consists in that, first, the quadratic objective function is block tridiagonal and, second, the inequality constraints are imposed individually upon each vector variable. The two nonstationary regression estimation algorithms considered in this paper are built as those of pair-wise separable quadratic programming and have the lineart computational complexity of the quadratic programming problem of general kind. The asymptotically strict iterative algorithm is based on the traditioal steepest descent method of quadratic programming, whereas the fast approximate algorithm consists in a single run of a special version of the dynamic programming procedure.

Important Links:

Go Back