Study On Blind Reconnaissance Technology For Communication Based On Neural Network

With the continuous development of communication technology, the complexity its environment and the ceaseless retread of signal modulation ways, the traditional communication counterwork systems is hard to meet the necessity of communication countermeasures. In this situation, the concept of blind reconnaissance is put forward with the purpose of establishing a set of reconnaissance theories for all the signals.With the leading concept of blind reconnaissance, the blind reconnaissance system for communication counterwork is designed which is comprised of three parts: instantaneous parameter extraction, feature vector extraction and feature matching. The technique can make reconnaissance without any prior knowledge, ignoring all the processes of parameters measurement, analysis and identification. Also the approach works overriding the signal format. It can guarantee the analysis and identification of all kinds of communication signals. The paper presents the method in detail, furthermore, the algorithm and implementation of signal instantaneous parameter extraction, feature vector extraction and feature matching. In the implementation, precise extraction of signal instantaneous parameter is achieved by wavelet transform. As the large quantity of data for instantaneous parameter, singular value decomposition (SVD) algorithm is implemented to reduce the quantity of data for instantaneous parameter, and saving the signal information to the largest extent. So the extraction of feature vector is realized. And feature matching is implemented by error back propagation neural network (BPNN). The simulation results show that application of this approach can identify the modulating style of common communication signals with a higher recognition ratio. Furthermore, the simulation results obtained show that the results of classification are desirable in some ratio of signal to noise when this system is applied to the classification of MASK, MPSK or MFSK.