Rapid HF waveform identification

Recognition algorithms for five selected waveforms have been successfully developed, based on determining a correlation for each. For the serial methods correlations are based on known training sequences; for the FSK methods correlations are based on expected spectral peaks. Selection of waveform is based on the most successful correlation, taking into account the required correlation threshold for each waveform. Thresholds may be adjusted to trade off detection failure and false recognition error probabilities. The target for recognition in about 0.2 second has been achieved. A rapid prototyping process has been used successfully to produce a system operating in real time on a multiple DSP system. Results have been produced by simulation, by using bench tests with real modems and a channel simulator, and by use of signals recorded over the air. The simulation and bench test results showed comparable and good performance for cases of interest, some examples of which are given in this paper. Results for the recorded signals were in line with expectations and required performance.