Higher-Order Spectral Processing and Wavelet Based Cumulant Processing

摘要

The objective of this program is to combine higher order spectral processing andwavelet processing to improve the detection and classification of certain types of signals that arise in sonar full spectrum processing and in condition based maintenance. We are focusing initially on two types of signals: coherently related narrowband tonals, and metallic transients. Coherent sets of narrowband tonals arise frequently from various types of rotating machinery. We are trying to exploit the coherent relationship between these tonals, as well as their proportional bandwidth properties, to improve the detection of these sets of tonals in noise and to provide additional information that can be used for condition based maintenance algorithms. Higher order spectral processing is used to exploit the coherence, and wavelet processing is used to exploit the proportional bandwidth properties.