Reformation Of The Undersampled Differential Frequency Signal Based On Deep Learning

According to the Nyquist sampling theorem,to retain all the information in the frequency component,the sampling rate must be twice or more than the highest frequency of the signal,and in practical environments,it is usually necessary to set the sampling rate higher to achieve accurate estimation of frequency parameters.At the same time,the large amount of data obtained at the high sampling rate means that the real-time computing amount and storage amount of signal processing will be very large,which puts forward higher performance requirements for hardware systems such as collection and storage.For the above problems,the paper aims to achieve signal reconstruction below Nyquist sampling rate,and takes the differential frequency signal used in linear FM radar system as the recovery object.We present a network model for signal reconstruction under 5 times under-sampling based on convolutional neural networks,and design a signal analysis software system for parametric measurement and spectral analysis of the reconstructed signals.Based on the theory of deep learning and Windows application design,this paper carries out the research of under-sampling signal reconstruction and parameter analysis.The main work is as follows:(3)Build a convolutional neural network model for reconstructing under-sampling signals: First,MATLAB software is used to create a differential frequency signal data set containing location and speed information;Secondly,the structure of the network model is determined through multiple sets of experiments,and the signal data set is used to train the network repeatedly,to obtain the optimal solution of the weight parameters in the whole network;Finally,the prediction value and the error are evaluated and analyzed.The experimental results show that the proposed method can reconstruct the original signal and extract the target position and speed information in the signal with a low error,which effectively reduces the sampling pressure of the signal acquisition module and plays an important role in improving the signal processing efficiency.(4)Application design for parameter analysis of signals: First,analyze the function and performance requirements of the software,including the measurement of each parameter,spectrum analysis,etc.,and limit the design objectives to define the basic functions and performance requirements of each module;Secondly,Windows Presentation Foundation technology is used on the.Net Framework development platform to realize the control layout of the main interface,and the detailed parameters,spectrum analysis algorithm writing is realized through C# programming language;Finally,run the software to test the correctness of each functional module.The operation process and results show that the functional modules of the software operate stably,and the accuracy of parameter measurement also reaches the expected goal.