Feature Abstraction And Automatic Recognition Of Modulation Schemes Under Interferential Environment

The so-called automatic recognition of modulation (AMR) schemes automatically classifies the modulation types from the samples of received signal by abstracting the related features such as amplitude, frequency, and phase, followed by employing proper recognition algorithms based on the fast digital signal processing techniques and advanced computing technology. The result of AMR can be used in the following processing blocks such as demodulation, supervision processing and anti-interference processing and so on. The potential applications of AMR include both civil and military communication, especially non-cooperative communications and communication confrontation, such as identifying signals, supervising signals, distinguishing interference, electronic confrontation, analyzing military threat, etc.On the basis of our analysis to the existing research on feature abstraction, the related feature abstraction methods are optimized in this paper, resulting several effective methods such as the feature abstraction based on transformation domain, stepped voltage level analysis, normalized carrier-free spectral energy analysis, squared signal and fourth powered signal analysis, etc. Both the decision theory based on recognition algorithms and the artificial neural network (ANN) based on recognition algorithms is analyzed, and the former is selected as it is more appropriate for this research.Under the condition that the carrier frequency is the only prior knowledge, various feature abstraction methods and AMR algorithms are analyzed, integrated and compared for the AWGN and non-AWGN channels. For 10 common modulation schemes, that is, AM, SSB, FM, CW, OOK, PSK, QPSK, FSK, MSK and 16QAM, the related algorithms are designed and simulated, and the corresponding successful recognition rate and the average recognition time for each algorithm are calculated. It is shown from our simulation that, when SNR is no less than 15dB, for AWGN environment, the correct recognition rate is higher than 98% and the average recognition time is about 3.2s; for Rican environment, the correct recognition rate is higher than 95% and the average recognition time is about 3.26s; for Rican environment, the correct recognition rate is higher than 96% and the recognition time is about 3.28s. In addition, based on fast Fourier transform (FFT), an effective and reliable carrier frequency detection method has also been designed in this thesis.