Research On Digital Self Interference Cancellation Algorithm Based On Deep Learning

Electronic warfare has become an important form of warfare in modern warfare.With the increasing complexity and variability of electronic warfare,it requires more military spectrum to meet the information exchange needs of wireless tactical communication.The simultaneous transmission and reception system can improve spectral efficiency,but it inevitably introduces high-power self interference signals.In the simultaneous transmission and reception jammer system,in order to accurately obtain enemy radar signals for interference,it is necessary to cancel the strong self interference signals coupled to the receiving end.This article studies the strong correlation and nonlinear self interference signals in the simultaneous transmitter and receiver jammer system,proposes a deep neural network(DNN)based self interference cancellation method for intra pulse segmented interception interference,and designs a convolutional deep neural network(CDNN)based self interference cancellation method for self interference generated by inter pulse full pulse forwarding interference,And the effectiveness and feasibility of the two methods proposed in this article in processing strong correlation and nonlinear self interference signals generated in their corresponding different application scenarios were verified through simulation analysis.Firstly,this article analyzes the basic working principle of the simultaneous transmitter and receiver jammer,introduces the ways of interference generation in the system of the simultaneous transmitter and receiver jammer,models linear and nonlinear self interference signals under a single channel,and simulates and analyzes the applicable conditions and shortcomings of the traditional adaptive filtering Least Mean Square(LMS)algorithm.Secondly,in response to the problem of traditional LMS algorithms not having ideal cancellation performance in dealing with strong correlation and nonlinear self interference signals,a DNN network structure was designed to replace the LMS algorithm to eliminate self interference signals.By utilizing the advantage of deep networks that can fit nonlinearity well,simulation analysis was conducted on the self interference cancellation of this network structure under the condition of uncorrelation and strong correlation between the target signal and self interference signal,The simulation results showed good recovery of the target signal and achieved a high cancellation ratio.On this basis,a real-time interference method of segmented interception is proposed for the simultaneous transmitter and receiver jammer,which can directly interfere with enemy radar signals in the current pulse segment.The DNN network structure is used to simulate and eliminate the linear and nonlinear self interference signals generated by this interference method.The simulation results show that compared with the traditional LMS algorithm,the DNN network structure proposed in this paper has different bandwidth,modulation methods The self interference under different signal-to-interference ratios,different forms of self interference,and different interference styles has good cancellation effects.In addition,the network model has also shown good cancellation ability in measured data.Finally,considering the problem of high network parameters when dealing with long sequences in the DNN network structure and the problem of self interference between different pulses of the transmitter and receiver simultaneously interfering machines,a CDNN network structure is proposed to eliminate self interference signals.First,convolution and pooling are used to compress the network input to reduce the sequence length of the input DNN network.The simulation results verify that the network structure has good cancellation results in the case of long inputs.At the same time,different ways of generating self interference between pulses were established,and based on this,a CDNN network structure was used to simulate and eliminate the generated self interference.The simulation results showed that the network structure can effectively suppress linear and nonlinear self interference signals.