Kybernetika

A novel approach to parallel transputer implementation of a regularized exponential parameter tracking is described. The proposed block regularized exponential algorithm allows to define and automatically adjust mean value of an alternative covariance of the estimated parameters in the boundaries of blocks of processed data. The alternative mean tracks in blocks the current parameter estimates. That is why the influence of the regularization on the parameter estimates is reduced, and the algorithm remains compatible with the completely pipelined parallel transputer implementation.