Signal estimation over short data records : a data-dependent time-invariant algorithm
Date of Original Version
A new algorithm for the estimation of stationary random signals from short data records consisting of signal plus noise is given. Time-invariant algorithms to solve this problem suffer from transient effects at the beginning and end of the data record, while optimal algorithms are time-varying and thus computationally expensive. The approach presented here consists of a time-invariant Kalman smoothing estimator modified by a data-dependent correction term involving an estimated initial state vector. A theoretical performance analysis of the algorithm is presented, and comparisons are made with the optimal Kalman smoother, as well as with a different approach based on Wiener smoothing. The results show that the proposed approach has nearly optimal performance and is computationally efficient. © 1990.
Journal of the Franklin Institute
Vaccaro, Richard J., and Fu Li. "Signal estimation over short data records : a data-dependent time-invariant algorithm." Journal of the Franklin Institute 327, 3 (1990): 439-455. doi:10.1016/0016-0032(90)90008-7.