Research On Communication Signal Detection And Recognition Technology Based On Parallel Computing

In the field of radio communication,signal detection and recognition is an important means of radio monitoring.As a non-cooperative party,it is often necessary to complete signal identification without prior knowledge.The difficulty of signal detection and recognition lies in the requirements of real-time,portability,and accuracy.Although traditional FPGA and DSP platforms can meet the above requirements,they rely on the selection of expert features,and the algorithms for blind detection of multiple signals are limited.Based on the above problems,this thesis proposes a detection and recognition algorithm based on the CPU+GPU heterogeneous platform.Apart from separating multiple signals,it not only meets the requirements of real-time and portability but also meets the requirements of the huge computing resources required by the modulation recognition algorithm based on deep learning.In this thesis,the communication signal detection and parameter estimation algorithm,modulation recognition algorithm and parallel computing method are deeply studied.The main contributions are as follows:(1)The detection method and parameter estimation method of communication signal are analyzed.In view of the shortcomings of existing modulation recognition methods that cannot solve the problem under the situation of in-band multi-signal,this thesis proposes a communication signal detection and parameter estimation algorithm suitable for multi-signal,and the algorithm is experimented.The results show that the algorithm can complete the detection and separation of the signal well,which can pave the way for the subsequent modulation recognition.(2)The common modulation methods of communication signals are analyzed,and the deep learning algorithm model is studied.According to the data characteristics of the communication signal,this thesis proposes a lightweight network model,and compares the recognition accuracy results obtained by using the signal time-frequency domain data and multi-dimensional feature data for model training.(3)In response to the real-time and portability requirements of signal detection and recognition algorithms,this thesis proposes a parallel computing-based communication signal detection and recognition algorithm on the basis of studying the GPU hardware architecture and CUDA programming model.The realization of the algorithm is completed by using Jetson TX2 platform.The algorithm greatly improves the recognition efficiency under the condition of ensuring the recognition accuracy,and can improve the detection accuracy of low signal-to-noise ratio signals,and reducing the cost of system improvement.