| Fast search technology of satellite communication signal under low signal-to-noise ratio is a hot research topic at present,which mainly involves signal detection,modulation recognition,signal acquisition and storage,signal capture and tracking and so on.It has great significance in military and civil satellite signal search field.The existing satellite communication signal search algorithms have some problems,such as large computation amount,poor signal feature extraction ability and weak signal decision ability,so it is difficult to have a good performance in both detection rate and real-time performance.In general,satellite communication signals are sudden in the time domain,and the signal detectors have enough template accumulation for analyzing signals.Aiming at the application scenario mentioned above,the research is carried out from the aspects of signal feature representation and signal rapid detection technology,and so the rapid search task of satellite communication signals under low signal-to-noise ratio is completed.The main work includes the following three aspects:1.In response to the problem of low computational efficiency in search algorithms for satellite communication signals,this paper conducts research on a fast detection algorithm for satellite communication signals based on perceptual hash feature representation.Firstly,the signal segmentation localization technique based on spectral entropy method is analyzed,and the robustness of spectral entropy method is verified.Then the effectiveness and noise reduction performance of Welch spectrum estimation method based on continuous wavelet transform are analyzed at the highest signal-to-noise ratio scale.Finally,the wavelet spectrum is represented by the subband variance sensing hash method,and the hash matching is carried out to complete the fast search of signals.It is proved that the perceptual hash method is effective and robust in the signal feature representation,and has a lower computational cost than other algorithms.Experimental results show that the proposed algorithm achieves good performance in both detection rate and real-time performance.2.In response to the problem of low recognition rate of fast search algorithms for satellite communication signals,this paper conducts research on fast signal modulation recognition algorithms based on graph convolutional networks.Firstly,the wavelet transform multi-resolution analysis theory is applied to decompose the signal into 10-level wavelet transform detail signal and construct the graph signal.The graph signal represents the information of the original signal in different frequency bands,which is a high dimensional characteristic representation of the signal.Then,neural network modules are selected to build a lightweight graph neural network model.The graph convolution layer in the model completes the fusion of graph signals and adjacency matrix,thus integrating the correlation between wavelet detail signals into the graph.Then,the graph attention layer strengthens the feature expression of graph signals and completes signal modulation recognition.The problem of gradient explosion and gradient disappearance is avoided,and good performance is achieved in both recognition rate and convergence speed.3.To meet the requirements of miniaturization and intelligence of the satellite communication signal search equipment,this paper selects edge computing equipment Jetson Nano as the hardware platform.A satellite communication signal search terminal software system is designed,which integrates the above algorithms and is deployed on the equipment.The detailed demand analysis is carried out in the system design process,the appropriate edge computing platform is selected,and the software design index is defined.After system deployment and debugging of the terminal,a good signal search effect is obtained in practical application,which proves the effectiveness of the proposed algorithm. |
PDF Full Text Request