Modulation Recognition And High Dynamic Synchronization Based On Cyclic Spectral Density

Modulation recognition and subsequent high dynamic synchronization are the technical bases of intelligent receiver based on software radio, and are also its bottleneck and kernel. They play an important role both in multistandard communications and also in signal scrutiny. Especially in recent ten years, computer technology, high speed digital signal processing technology as well as high speed special devices ensure the engineering realization of modulation recognition and high dynamic synchronization technologies, and also its important status in noncooperative communication and its applied value for software radio and cognitive radio are much more understood .Their essentialities are recognized by more and more researchers.In the background above, the problem about modulation recognition and high dynamic synchronization are investigated by virtue of cyclic spectral correlation theory in this paper. The research contents include amelioration and realization of cyclic spectral correlation algorithm, modulation recognition algorithm and high dynamic synchronization based on cyclic spectral correlation.Firstly, for cyclic spectral correlation algorithm, the signals can not be processed real-time because of its high computing complication. So, the modified algorithm of cyclic spectral correlation is presented. When its resolution was not integral of power of 2, DFT was computed employing sliding FFT algorithm; otherwise, DFT was calculated using AFT; and correlation was processed using one bit correlation algorithm. Number of addition was chosen as the weigh criterion of computing complexity, and then the closed expression of computational complication of presented algorithm was given by virtue of conversion parameter between real addition and real multiplication, and performance was simulated. Subsequently, the problem about realization of algorithm was resolved in terms of soft kernel processor NiosII and its Avalon bus configuration.Taking SSCA as an example, the realization process and method for the realization of CSD were given.Secondly, aiming at deficiencies of present modulation recognition algorithms and different applied background, ameliorations were made in both feature parameter and classifier according to process of statistical modulation recognition method. For feature parameter, some ill-suited cyclic spectral correlation parameters were removed and other suited parameters were added. Simultaneously, some frequency domain feature parameters were gained in the process of computing CSD according to the characteristic of CSD parameter. So recognition probability and the modulation type were increased without increasing computing complication. For classifier, at first, probabilistic neural network was employed, and its structure was modified. Comparing to conventional BP neural network, recognition time was decreased and successful recognition probability was improved. After that, the idea is presented using the self-organizing feature map neural network (SOM) as the classifier of modulation recognition,which exploited its properties of self-organizing,unsupervised and self-adapted to accommodate automatically the varieties of the signal to noise ratio. In order to enhance the recognition performance and decrease the recognition time, its learning rule and competitive transferring function were modified. The simulation results proved that recognition probability of modified SOM is higher than other neural networks. In addition, new methods were exploited to recognize modulation type, which used magnitude in bifrequency plane of CSD and applying the principle of minimum value of the sum of error-square to recognize modulation type. The algorithm could furthest utilize information embodied by CSD to improve the recognition probability. In the following section, the problem on modulation recognition of multisignal in the same frequency band was researched, for which a mathematical model is given. Combined with the theory of cyclic spectral correlation, modulation recognition algorithm for multisignal in the same frequency band based on cyclic spectral correlation is presented in terms of the presented mathematical model.Finally, high dynamic synchronization was researched using DSSS-QPSK signal. First of all, traditional costas loop and AFC was modified.Firstly, working fashion of them was modified, that is meaning AFC loop and costas loop work synchronously, so frequency difference and phase difference were eliminated simultaneously.Secondly, obtained frequency difference and phase difference were updated using sliding algorithm, which shortened the update time of them. In this way, loop could be lock quickly, contradiction between tracking precision and loop was resolved basically. The modified algorithm provides a comparision target for the following algorithms for high dynamic synchronization. In the following section, according to the anti-interface and anti-noise properties of CSD, high dynamic synchronization algorithm based on CSD is presented. Obtained in-phase signal throw costas loop was normalized to obtain a cosine signal. CSD of this cosine signal of was computed to obtain frequency difference and phase difference and then results were put into NCO to synchronize. The simulation results proved the exactness and the superiority of the presented algorithm. And then the performance of presented algorithms was compared, it is found that performance of algorithm based on CSD is better than algorithm based on AFC, but its computation is complicated.