Efficient Time-Recursive Implementation of Matched Filterbank Spectral Estimators

摘要

In this paper, we present a computationally efficient sliding window time updating of the Capon and amplitude and phase estimation (APES) matched filterbank spectral estimators based on the time-variant displacement structure of the data co-variance matrix. The presented algorithm forms a natural extension of the most computationally efficient algorithm to date, and offers a significant computational gain as compared to the computational complexity associated with the batch re-evaluation of the spectral estimates for each time-update. Furthermore, via simulations, the algorithm is found to be numerically superior to the time-updated spectral estimate formed from directly updating the data covariance matrix.