Research On Spatial Spectrum Sensing Algorithm Based On SVM Via Beamspace Transformation

The popularization of 5G technology has brought the demand for higher throughput and transmission rate,thus making the initially scarce radio spectrum resources more valuable.However,the traditional static allocation strategies limit the efficient use of spectrum,causing the spectrum crisis to become more and more dangerous.In order to alleviate the spectrum crisis,cognitive radio technology based on spectrum multiplexing is proposed.The core of spectrum multiplexing is to detect the surrounding idle spectrum through spectrum sensing technology so as to provide secondary users with dynamic access opportunities.Therefore,spectrum sensing,as an essential means to solve the multiplexing problem and optimize the spectrum efficiency,will play an essential role in the future wireless communication system.Further research found that realizing spectrum multiplexing from the spatial dimension can enable primary users and secondary users to coexist at the same time and the same frequency,further improving the efficiency of spectrum multiplexing.On the other hand,today's communication scenarios tend to be diversified gradually,and the traditional fixed decision threshold spectrum sensing scheme is difficult to flexibly adapt to the changing wireless environment while maintaining excellent detection performance.In order to deal with the above problems,an improved scheme based on beamspace transformation is proposed to solve the problem of the high computational complexity of the previous Weighted Multi-signal Classification(WMUSIC)spatial spectrum sensing algorithm in large-scale array antennas.The improved scheme reduces the dimension of the matrix by constructing an appropriate beam transformation matrix and weighting the beam transformation matrix with the original received signal matrix,thus reducing the computational complexity in matrix decomposition.At the same time,the beamspace transformation has the ability of spatial filtering,which can effectively filter noise and interference outside the observation range and improve the output signal-to-noise ratio.The simulation results show that the beamspace transformation has accurate angular resolution and good robustness to out-of-band interference,thus improving the detection accuracy of the spectrum sensing system.Secondly,the traditional WMUSIC sensing algorithm does not have a clear decision threshold calculation method.Moreover,the fixed decision threshold cannot adapt to the dynamically changing wireless environment and is prone to misjudgment.Moreover,the fixed decision threshold cannot adapt to the dynamically changing wireless environment and is prone to misjudgment.Therefore,based on the excellent binary classification characteristics of support vector machines,we use the test statistic obtained by beam space transformation as features for model training,so that the presence or absence of primary user signals can be adaptively judged.Compared with the fixed threshold detection algorithm,the proposed improved scheme has the ability of self-learning,can adapt to the changing wireless environment and higher classification accuracy.Simulation results show that the proposed scheme outperforms other related multi-antenna sensing algorithms,especially under low signal to noise ratio and low snapshot.Finally,inspired by the WMUSIC algorithm,this paper proposes a joint spatial spectrum sensing scheme based on the covariance matrix by combining the spatial spectrum estimation algorithm on account of beamforming and array linear prediction.In this scheme,two spatial spectral functions are obtained by inverting the covariance of samples,and their respective spectral value ratios are used as test statistics.Then,the two statistics are combined as features for SVM model training,and the trained model is used to judge whether there is a primary user signal.This scheme can obtain the corresponding angle information while detecting the primary user,which is of considerable significance to the space multiplexing of the spectrum.The experimental simulations show that the proposed joint detection scheme has better detection performance than the WMUSIC algorithm,and the model using joint feature training is more robust than the model using single feature training.