Research And Platform Construction Of Emission Source Identification Based On Signal Fingernrint

With the advancement of signal processing technology,a large number of communication devices have emerged,leading to the complicated electromagnetic environment.Therefore,as one of the core problems in wireless networks,security issue should be given wide attention.In wireless networks security,the individual identification of communication transmitters is the vital target to solve.Based on the principle of transmitter signal information,this thesis studies the signal fingerprint in the power amplifier and the transmitter characteristics.This thesis establish the effective signal fingerprint model,feature extraction method,and neural network to complete the individual transmitter recognition.The main contributions of thesis are shown as follows:1)Based on the non-linear principle of the power amplifier,a deep memory polynomial model is established.The cyclic spectrum features extraction are performed on the received signal passing through the power amplifier.At this time,the cyclic spectrum map is obtained by super-imposing a plurality of cyclic spectrum maps with different symbol rates together.In this way,the cyclic spectrum map can extract more effective features.Then the cyclic spectrum features are sent to the multi-core path network to achieve the transmitter identification.Experiments are performed on Gaussian white noise channels and Rayleigh fading channels.Simulation results show that compared with Inception,ResNet and DensNet,the proposed multi-core path network is more robust.The experimental results show that when signal-to-noise is 10dB,the classification recognition rate can reach 97%under Gaussian white noise channels,while the result reaches about 92%under the Rayleigh fading channel.Simulation experiments prove that the proposed method can effectively achieve individual identification.2)This thesis designs a digital signal receiving model.The correct signal reception is conducted through simulation correction models such as automatic gain control,raised cosine matched filter,frequency correction,and phase correction.Then the universal software radio equipment USRP×310 is utilized to build an individual identification experiment platform.The signals are got in USRPx310 with the overall transmitter characteristics.Then the collect signals are sent to multi-core path network for identification.The actual measurement results show that the proposed network performs the best recognition accuracy for the target transmitter among Inception,ResNet and DenseNet.