Modulation Recognition Methods Of Digital Communication Signals

In order to complete the signal demodulation and other communication tasks, the signal modulation modes and modulation parameters must be identified correctly. Therefore, the automatic modulation recognition is very important. In this thesis, with the purpose of the better recognition performance under fading channel, the collaboration modulation recognition algorithm is improved based on the existing single node modulation recognition algorithm, combining information fusion algorithm.In order to reduce the impact of deep fading, shadowing, hidden nodes and other problems in wireless communication environment, this thesis presents a new cooperative recognition method to correctly identify 13 kinds of different modulation signals, which are 2ASK,4ASK,8ASK,2PSK, QPSK,8PSK,2FSK,4FSK,8FSK,16QAM,32QAM,64QAM and OFDM. First, the nodes in collaboration extract different types of feature parameters, these parameters are based on the instantaneous information, combination of wavelet decomposition detail coefficients and high order cumulants,and signal cepstrum coefficients, which can characterize the target signal from different aspects, then each node inputs the feature parameters to the trained BP neural network to have a test, then the output of the neural network as evidence directly sent to the fusion center to fusion using an improved D-S evidence theory. The simulation results show that the proposed fusion method not only can improve the average recognition perfonnance but also can increase the identifiable signal types.A two-stage cooperation modulation recognition algorithm based on the correlation clustering is proposed. In order to realize correctly recognition of several classic modulation types such as BPSK,2FSK,2ASK, QPSK,4FSK and 4ASK, the modulation signal is decomposed into 7 levels by the db3 wavelet at first, then the signal of every level is reconstructed, therefore, S1-S8 eight kinds of features are obtained through computing the signal’s average variance S, another four features are added so that the correct recognition rate of 2PSK, QPSK,2ASK can be improved, What is more, a two-stage cooperation modulation recognition algorithm clustering based on the correlation of the received signals is introduced to solve the well-known hidden user problem. Here, we organize the users in clusters based on the correlation of the energy of their received signals, the users in the same cluster take different feature according to its received signal’s SNR to cooperate, elementary feature fusion is carried on in the cluster head and decision-making fusion is carried on in the fusion center. The simulation results show that the proposed algorithm has higher recognition rates and better system reliability compared with the random clustering, the mean correct recognition rate at -15dB can reach 82.50%.Collaborative modulation recognition algorithm based on maximum likelihood decision theory is studied. First, from the perspective of the probability density function in the likelihood function expression, four simplified approximate probability density function is given:Tikhonov probability density function, L-order Fourier coefficient approximate probability density function, Gauss-Legendre finite integral approximate probability density function, the Gauss-Hermite semi-infinite integral approximate probability density function. The Kullback-Leibler distance of the four approximate probability density function between the exact probability density function is calculated respectively, used to quantify how close they are to the exact probability density function. Then a collaborative modulation recognition algorithm based on maximum likelihood decision theory is given, the Gauss-Hermite semi-infinite integral approximation probability density function which is closest to the exact probability density function is adopted in the likelihood function. The simulation results show that, the correct recognition probability compared with the single-sensor recognition performance, improved 1dB when 3 sensors collaboration,and 3dB when 10 sensors collaboration.