Noise with alpha stable distribution leads to loss of efficacy of the second-order cyclic spectrum coherence coefficient, and degrades related algorithms for communication signal modulation recognition. A recognition algorithm based on fractional low-order cyclic spectrum coherence coefficient is proposed to solve this problem. The related theory of fractional low-order cyclic spectrum coherence coefficient is first introduced.
Fractional low-order cyclic spectrum coherence coefficients of communication signals are analyzed. Based on the analysis, the algorithm extracts the cyclic frequency profile of spectrum coherence coefficient as the recognition characteristic parameter, and uses a BP neural network as a classifier to achieve communication signal modulation recognition. Simulation results show that, in an alpha stable distribution noise environment, the performance of the proposed algorithm is superior to that based on second-order cyclic spectrum coherence coefficient. These two algorithms have the same performance in Gaussian noise.