Blind Identification Of Digital Modulation Signals In Low Snr

In this dissertation, the approaches of digital communication signal modulation recognition are studied. We get the effective methods for the feature extraction by studying the characteristics of different modulation styles. The advantages and disadvantages of existing modulation recognition algorithms are analyzed. The performance of these algorithms at low SNR is also analyzed too. In order to achieve blind recognition system at low SNR, the variety of all the algorithms'applications are discussed in detail. Finally, a recognition system is designed by using different algorithms according to the characteristics of different modulation. Then, the system performance is validated through the simulation experiment. The main work are listed as follow:1, We use the high-order moment feature of the signal. Because of the difference in the constellation, in the second-order moment signals show different features which can be used to differ each other. Simulation shows that , in the condition of snr > 0dB the recognition rate can reach 90%; when snr < 0dB, the algorithm can not work.2, We use the signal's envelope spectrum feature and power spectrum feature, square spectrum feature, higher-order spectrum feature. Using these features, we achieved modulation classification of 2FSK, MSK, BPSK, QPSK, OQPSK signals. A modulated signal will show a peak in a certain spectrum, so we can make the numbers of these peak as characteristic parameters for modulation recognition. Simulation shows that, 2FSK, MSK, BPSK, QPSK signals can be classified correctly in the condition of snr > - 5dB; in the condition of snr > 3dB, the recognition rate of OQPSK is over 90%.3, We use the cyclic spectrum feature of signal. The distinguish between BPSK and UQPSK is a problem. They have similar features, so common classification algorithms are not available. Cyclic spectrum is the extension of power spectrum. In the cyclic spectrum, some features of signal become visible. BPSK and UQPSK perform different in the cyclic spectrum, so we can make it as characteristic parameter for modulation classification. Simulation shows that, BPSK and UQPSK can be classified correctly when snr > - 4dB.4, Recognition system is design by using the classification method mentioned above. With on priori knowledge, this system can a achieve the correct classification of 2FSK, MSK, BPSK, QPSK, OQPSK, UQPSK signals at low snr. In the condition of snr > - 4dB, the recognition rate of 2FSK, MSK, BPSK, QPSK, UQPSK can reach 90%; in the condition of snr > 3dB, the recognition of OQPSK can reach 90%.