A Study Of Digital Modulation Recognition Algorithm Of Communication Signals At Low Signal-noise Ratio

Communication signals modulation recognition technology has gone through years of development. Many scholars introduced various techniques to the field,and achieved fruitful results. Now the non-cooperative communication is still important forms of communication used in the military and civilian fields. The development of communication technology and the complexity of the wireless communication environment present new requirements for modulation recognition technology.This article focuses on the research of efficient recognition in low SNR circumstance.The main work of Modulation recognition is to preprocess the received signals,extract feature parameters, and design classifier to recognize the modulation. The main work and achievements are as follows:1. Improving signal recognition algorithm based on feature parameters extracted from instantaneous information of signals. The three new characteristic parameters extracted, respectively, are the maximum amplitude of Fourier transform of the absolute value of normalized Zero Center instantaneous amplitude, the number of the normalized instantaneous frequency which is greater than 20, the maximum amplitude of Fourier transform of non weak signal segment of the normalized instantaneous frequency.Simulation results show that the improved algorithm is better than traditional one and that based on wavelet transform.2. Studying the signal recognition algorithm based on feature parameters extracted from the power spectrum of signals,analyzing their estimation and combining with the clustering algorithm of the signal.The algorithm makes a recognition rate up to 99%when the signal-to-noise ratio greater than 5d B.3. Researching the signal recognition algorithm based on feature parameters extracted from the high order cumulant of signals. With the filtering of signals before the calculation of high order cumulant, using different methods for different signal to solve the high order cumulant, and modifying the high order cumulate of 4FSK signal which is not correctly described in related articles, the signal to noise ratio of the effective identification can be as low as 0d B.4. On the research of BP neural network, and by five characteristic parameters as the input of the network, the network is trained, the simulate results indicated that the combination of the five characteristic parameters and the neural network achieves high recognition rate in the low SNR.