Design And Realization Of A Method Of Automatic Modulation Recognition Of Digital Signals

Digital signal has a wide range of applications in the fields of communications, radar, etc, due to the merits of anti-interference ability and easy of encryption. Digital signal modulation recognition is an intermediate step between signal interception and demodulation, it has been an important issue in electronic warfare, spectrum management, and software-defined radio in the past two decades. Based on the previous studies, the automatically classification modulation of MPSK、MFSK and MPSK/MQAM is studied and improved in this thesis.As the higher order cumulants preserve the phase information of signals, it can characterize the distribution of the constellation very well. At the same time, since the higher order cumulants of AWGN is zero, it performs good noise immunity. In this thesis, the identification algorithms of MPSK and MPSK/QAM are both base on high order cumulants. After eliminating the effect of shaping filter to the amplitude of the MPSK signals by the normalization of envelope, the signal multiplying by its shifted-conjugated, the higher order cumulants becomes applicable in the intermediate frequency signal which can only be applied in the baseband signal. After the analysis the best shift points are given. According to the nature of the higher order cumulants, two parameters are constructed to achieve the MPSK signal classification. The signal which is not normalized by envelope is also multiplied by its shifted-conjugated, then calculate its higher order cumulants and construct two classification parameters, then the MPSK and QAM can be distinguished. Simulation results show that the algorithm of this thesis has higher recognition rate and shorter response time than other commonly algorithms, such as instantaneous phase method, spectral analysis, the cyclic spectrum method and higher order cumulants method.For the MFSK signal modulation recognition, the algorithm of higher order cumulants combination with spectral analysis is adopted base on the instantaneous frequency in this thesis. The algorithm can correctly identify the signal modulation within a certain range of SNR and code rate. Compared to the standard deviation of the absolute value of the instantaneous frequency, the number of peaks, higher order cumulants of signal differential and other analytical methods, this algorithm can achieve higher recognition rate just by increasing a little computing. In addition, it owns the ability to achieve MFSK signal modulation recognition of non-integer ratio of symbol rate to the center frequencyFinally, a hardware system is introduced, and the algorithms are implemented on this system as well. With the offline trained support vector machines written into the DSP, the accuracy of classification is higher than90%with center frequency is160MHZ and SNR higher than a certain value (phase modulation signal is higher than6dB, the rest is higher than8dB). Finally, the commonly used algorithms are compared with the proposed algorithm in terms of the recognition rate and response time on the DSP chip. The hardware system further proves the achievability, superiority and effectiveness of the proposed algorithm.