Research And Implementation Of Communication Signal Modulation Recognition Technology Based On Feature Extraction

In modern communication systems,modulation recognition plays an extremely important role in both military and civilian applications.In the military field,modulation recognition must be realized in order to intercept the message of the enemy.In the civilian field,the monitoring of the spectrum is also inseparable from modulation recognition.With the rapid development of communication technology,modulation methods have become more diverse,and the electromagnetic environment has become more complex,and there are more interference sources in the transmission process,which has placed higher demands on modulation recognition technology.Therefore,it is of great significance to study the modulation recognition technology.In this thesis,the recognition theory based on time and frequency domain features,high-order cumulant features,and constellation features are studied.By combining different theories,the identification scheme is designed and simulations are completed.16 communication signals are modulated and identified,involving 10 different eigenvalues.In the designed scheme,the inter-class identification of communication signals is first performed,and when the signal is a digital modulated signal,intra-class identification is executed.In the intra-class identification of FSK signals,the frequency domain feature extraction is carried out,and a statistical algorithm of the number of peaks is proposed.Compared with the FSK recognition algorithm based on time domain features,the recognition rate is significantly improved.Clustering is used in the identification process of QAM signals.When there is phase difference between the mixed signal and the received signal during the down-converted process,the constellation diagram rotates,resulting in recognition error.In this thesis,a constellation rotation degree calculation method is proposed for this problem.By calculating the constellation rotation degree,the clust radius is corrected,and improves the recognition rate of the QAM signals significantly.Another commonly method for calculating the rotation degree of a constellation is based on Radon transform.Compared with this algorithm,the proposed algorithm has lower computational complexity and higher accuracy under certain SNR conditions.Finally,the designed scheme is implemented.The overall structure of the module,and the detailed design of each sub-module are given.Based on the existing results of the project team,a test platform is built and the modulation recognition module is tested.For the FM modulation,a real broadcast signal is received and the recognition rate is above 92%.For the five modulation methods such as 2ASK,the tested signals come from a signal generator.The output power is-80 dBm,and the recognition rate is greater than 92%.For the other ten kinds of signals such as 4ASK,since the signal generator does not provide these modulation methods,the tested data is generated by Matlab and superimposed with noise,and the recognition rate is greater than 93% under the condition of SNR=13dB.It can be seen from the test results that the design and implementation in this thesis have achieved good results.