Sweep operator applied to the calculation of partial and multiple coherence

Abstract

In this work, a transformation to a square matrix named sweep operator is proposed to obtain multiple and partial coherence estimates. This operator does not require the inversion of the spectral matrices commonly involved in coherence estimation and hence may be useful in applications such as in embedded systems, where memory use is a critical issue. The technique was evaluated in electroencephalographic data and its performance was compared to that using the conventional approach. The results indicate the sweep operator to be a promising alternative in applications where the memory use is a crucial issue.