| Individual emitter identification, also known as the radiation fingerprint identification, is an important issue in the field of communication reconnaissance in recent years. It refers to the process of feature extraction from the received communication signals and then determining the individual emitter that generating signals according to the priori information. Therefore, analyzing subtle features of emitter signals and extracting the different features from other emitter signals is very important in the process of individual emitter identification.In this paper, we discuss several different feature extraction methods of local integral bispectrum, then we compare the advantages and disadvantages of several existing local integral bispectrum and choosing bispectrum in experiment. We mainly focus on extracting the signal features with the square integral bispectrum method. The reasons are as follows. On one hand, the square integral bispectrum has the features of time shift invariance and scale transformation and phase stability, on the other hand, it can sample the bispectrum without omitting and repeating. We take the improved method of largest proportion interval on the basis of the square integral bispectrum to remove low-contribution and negative-effect bispectrum values, moreover, redundancy and noise brought by this part can be reduced. Experimental results show that the improved method can improve recognition rate. What’s more, the improved method also has good recognition rate under different SNRs. We adopt the method of analyzing principal components of adaptive combined kernel function. The kernel function can give consideration to both global features and local features, therefore, it can reduce the dimensions of characteristic vector. The experiment results verify that the efficiency of computing can be improved greatly on the condition of ensuring the rate of identification when this method is used for the reduction of feature vector dimensions. |
PDF Full Text Request