| Big Doppler frequency shift is a remarkable character of the low orbit satellite communications, which will cause the deterioration of output SNR of the acquisition circuit result from Doppler frequency. Auto frequency control (AFC) loop is introduced in known Doppler frequency shift compensate scheme, but big Doppler frequency shift greatly exceed the estimate range, reach about±100kHz. Traditionally big Doppler frequency shift is divided into some smaller areas, and serial or parallel technique for searching frequency child area is adopted. Because the searching time of the serial technique is very long, and the parallel technique will expend a lot of hardware resource, the performance of the communication system will be affected.The main achievements of this dissertation are:1. The Doppler frequency child area searching technique basing on synergetic neural network is proposed. The different frequency areas can be regard as different patterns, and synergetic NN has the particular advantage in pattern identifying, which lead to this technique is feasible.2. The simple approximate form of adjoint is deduced by analyzing. And on the condition of lacking transcendent knowledge, the probability of the Doppler shift on each area is equivalent. The initialization order parameter is necessary. Accordingly the complexity of this arithmetic becomes small.3. Compare the technique of searching frequency basing on synergetic with serial or parallel technique, analyze and summarize the character of different arithmetic, testify the preponderance of this scheme from the complicacy.4. Estimate the frequency basing on BP, RBF, GRNNS with a lot of simulations.Research results indicate:1. The synergetic neural network technique can find the Doppler frequency child area quickly, and consume less resource. So this scheme can compensate the Doppler shift effectively.2. BP network can not reach the high precision, and the size of RBF, GRNNS network is very large, which would be adverse to engineering.
|
PDF Full Text Request