Research On Modulation Classification Algorithm Of MIMO Communication System Based On Distribution Test

Communication modulation mode has evolved from the earliest analog modulation modes such as amplitude modulation and frequency modulation to multi-order digital modulation modes,providing extremely flexible signal transmission methods for different communication systems.Judging from the main research achievements of domestic and foreign scholars in the modulation and recognition of MIMO systems in recent years,there are mainly the following two problems:(1)The modulation recognition algorithm based on likelihood is difficult to solve the calculation complexity problem in MIMO systems.With the increase of the signal modulation order,the computational complexity increases exponentially;(2)The recognition algorithm based on traditional signal characteristics requires accurate channel information and lacks practical application scenarios feasibilityIn view of the high computational complexity of the modulation recognition algorithm based on likelihood,this thesis proposes a modulation recognition method based on distribution testing.By comparing the empirical cumulative distribution function of the observed signal with the theoretical cumulative distribution function of each candidate modulation mode,the candidate modulation mode with the highest fitting goodness is selected as the result of modulation recognition decision.The simulation results show that the distributed test can obtain the recognition performance matched with the ML classifier when the signal-to-noise ratio is high,but has a lower algorithm complexity.To deal with the problem that the three distribution tests will magnify the mismatch when the signal is not properly equalized or the samples are not evenly distributed among all modulation symbols,we propose a classification mechanism that uses a multi-layer perceptron to achieve the integration of Var test and multi-distribution test.The simulation results show that under different channel conditions,the DTE classifier proposed in this thesis is superior to the classifier based on a single distribution test,and when appropriate signal components are selected for analysis,DTE can provide excellent robustness in fading channels.In the non-cooperative communication channel environment,the signal receiving end can only directly obtain very little channel state information.We propose to use the expectation maximization algorithm to estimate multiple channel parameters to implement a blind modulation recognition classifier based on distributed testing.The simulation results show that the distribution test after EM parameter estimation can obtain the recognition performance consistent with the classifier of known channel parameters under the condition of high signal-to-noise ratioIn summary,applying distributed testing to modulation recognition is a feasible method,and has better recognition performance and stronger applicability.