Improving Transient Signal Synthesis Through Noise Modeling and Noise Removal

摘要

This thesis examines signal modeling techniques and their application to ambientocean noise for purposes of noise removal and for generating realistic synthetic noise to add to synthetically generated transient signals. Higher order statistics of the noise are examined to test for Gaussianity. Stochastic approaches to AR, MA, and ARMA modeling are compared to see which technique yields the best synthetic noise. Results from the modeling process are used to develop a short-time Wiener filter which can be used to condition a real signal for further processing through effective noise removal.