Adaptive erasure generation for coded frequency-hop communications with partial-band interference and fading

The erasure of unreliable symbols can significantly improve the performance of frequency-hop communication systems. In this paper, Bayesian decision theory and parameter estimation techniques are used to decide which symbols to erase. Erasure decisions are made by combining the envelope detector outputs into statistics that estimate signal-to-noise ratios and symbol reliabilities. With these techniques, no a priori channel information is needed and no test symbols are required. The performance of these techniques in frequency hop systems with Reed-Solomon coding is determined for channels with partial band interference, fading and wideband noise. This system provides performance comparable to the theoretical performance that would be obtained if perfect a priori channel information were present. In addition, performance is much superior to that obtained by errors-only decoding.