Partial Forward-Backward Averaging for Enhanced Frequency Estimation of Real X-texture Modes

Abstract:
In this paper, enhancement of the signal root estimation of a particular kind of real-valued two-dimensional (2-D) sinusoidal modes is considered. To its constitution, each mode corresponds to the superposition of two real-valued plane waves in a particular symmetry. The concept of partial forward-backward averaging, which is applicable for modes that are undamped in at least one dimension, is introduced as a means for improving the signal subspace estimate from which the signal roots are estimated. The consequences of real-valued signals for the signal root estimates are discussed in detail, and it is shown that by applying partial forward-backward averaging, the mean square errors of the estimates, and the breakdown threshold signal-to-noise ratio (SNR), are significantly reduced, compared with forward-only or conventional forward-backward (when applicable) usage of the sampled signals. The practical implication is highlighted by applying the proposed technique to modal analysis of multichannel impact responses from a tree trunk.