A Neural Network Based Levinson-Durbin Method For Adaptive Active Sensor Waveform Synthesis

Adaptive waveform design in active sensors has proved to have a tremendous effect on their performance. One of the most important parameters of an active sensor is its probability of detection, that its derivation becomes greatly complicated in signal-dependent backgrounds. In this paper a novel method for inter-pulse code synthesis is introduced that outperforms the reported signal design methods for detection of point targets in Gaussian stationary clutters while having nearly the same complexity. This method called ANN-based Levinson-Durbin procedure mitigates the need for large-order models and is robust to the truncation errors in designed waveforms due to limited time on target and number of pulses available in each CPI.