Transient high frequency signal estimation: A model-based processing approach

By utilizing the superposition property of linear systems a method of estimating the incident signal from reflective nondispersive data is developed. One of the basic merits of this approach is that, the reflections were removed by direct application of a Wiener type estimation algorithm, after the appropriate input was synthesized. The structure of the nondispersive signal model is well documented, and thus its credence is established. We merely state the model and devote more effort to practical methods of estimating the model parameters. Though wealth of efficient methods exist, for calculating the reflection weights, a simpler approach was employed here, since our reflective signal model allows us to make some basic simplyfying assumptions. The technique essentially consists of calculating ratios of the autocorrelation function at lag zero and that lag where the incident and first reflection coincide. We initially performed our processing procedure on a measurement of a single signal. Multiple application of the processing procedure was required when we applied the reflection removal technique on a measurement containing information from the interaction of two physical phenomena. All processing was performed using SIG, an interactive signal processing package [6]. One of the many consequences of using SIG was that repetitive operations were, for the most part, automated. A custom menu was designed to perform the deconvolution process.