| dc.description.abstract | Knowing the waveform or pulse-compression technique of a radar is a very useful feature in Electronic Support Measures (ESM) systems. Classification of radars (and their platforms), deinterleaving radar pulses, and efficient/smart jamming are some purposes. As pulse-compression becomes more advanced, and the peak power decreases in favour of spreading the energy over time, classification of the waveform becomes more difficult as the diversity increases. Several different phase-modulations and frequency modulations exists and are in use, e.g. Linear FM Chirp, biphase and polyphase codes and Costas frequency codes. In this thesis an automatic modulation recognizer (AMR), originally developed for communication signals by E. E. Azzouz and A. K. Nandi, has been implemented in LabVIEW and adapted to classify radar intrapulse modulations. Testing showed that the algorithm successfully classified unmodulated-, LFM-, PSK2-coded and PSK4-coded pulses down to 7dB SNR, and it is believed that even better results can be achieved by further tweaking of the AMR. The bandwidth of the receiver had a large influence on the key-features that were extracted by the algorithm. The AMR still needs to be tested on real radar signals. It is necessary to first develop a working carrier-frequency estimator and investigate the influence of inaccurate frequency estimation on the extracted key-features. Azzouz & Nandi also proved artificial neural networks (ANN) superior to the decision tree as classifier, and future implementations should by that reason use ANN's instead. ANN's were avoided here as time did not permit implementation. | nb_NO |
