Maximum Likelihood Adaptive Neural Systems (MLANS) Application to High Frequency(HF) Propagation

摘要

The feasibility of applying a model-based neural network technique to investigatethe properties of ionospheric clutter observed in the operation of high frequency (HF) propagation systems was examined. Individual ionospheric clutter structures found in the amplitude-range-Doppler (ARD) spectra of over-the-horizon (OTH) radar data were successfully segmented and characterized. A multi-mode Gaussian clutter model was formulated using the Maximum Likelihood Adaptive Neural System (MLANS) to fit the observations. The results indicate that either a three or a four mode Gaussian model is sufficient for MLANS to segment and characterize the observed clutter. High Fidelity simulations of time slices of the raw data were achieved by combining time-varying Gaussian together with a time-varying uniform distribution to represent the noise floor. Each Gaussian mode (or model) is characterized by a time-varying set of three parameters: amplitude, Doppler spread, and Doppler shift.