Shortwave Channel State Feature Extraction Andstate Prediction

As the transmission medium of the signal,channel is very important to effectively analyze channel characteristics and establish channel model in communication to improve the communication quality and achieve the best reception.Compared with other communication channels,the wireless channel has the most changes.Different transmission methods will form different channel models.Therefore,the analysis of wireless channels is very complicated.Traditional wireless channel analysis methods mainly include classical channel estimation algorithms such as MUSIC,channel impulse response analysis,and principal component analysis.The main problem of the traditional method is that it has limitations,is not comprehensive enough,and fails to verify the effectiveness of the method.It is limited by the channel model itself and the calculation is complicated.The new wireless channel analysis methods mainly include change domain analysis and imaging analysis.The change domain analysis method is highly adaptable,and the influence of external interference on the algorithm is small,but most of them use the received signal as the source,which is not enough to analyze the real-time channel characteristics.Therefore,in order to solve the problem of insufficient channel characteristics in the analysis of the change domain,this paper uses the multipath Rayleigh fading channel built by the Jakes model to obtain the channel coefficients in the slow fading state through deconvolution,.so as to obtain the 1000 signal samples in a certain error interval.The channel gain matrix is formed,and then the channel state image under the change domain is obtained.In order to realize the subsequent channel state prediction,through MATLAB simulation experiment,combined with the bit error rate interval,the channel state is roughly divided into three states: good,medium and poor.Then,the real-time channel state information is used as a data source for time domain,frequency domain,and correlation domain visualization.Corresponding contour maps,waterfall maps,and correlation maps are obtained,and three types of channel state images under the change domain are initially obtained.In view of the strong structural characteristics of the obtained state images,this paper proposes a multi-structure feature extraction algorithm.Under ideal conditions,the simulation results show that MSC has a high channel state prediction recognition rate,and all three change domains can reach Above 80%,the recognition effect of related domains can reach 90%.Secondly,in order to reduce feature redundancy and improve algorithm performance,PCA ?MDS?KPCA and Isomap are used to reduce the dimensionality of the aforementioned channel state images.Experiments show that the PCA dimensionality reduction algorithm is most suitable for the feature dimensionality reduction of this article.The time domain data selects the 21-dimensional features under PCA,the frequency domain data selects the12-dimensional features under PCA,and the correlation domain data selects the 21-dimensional features under PCA.Finally,the robustness of the algorithm is analyzed in terms of rotation invariance,resistance to light intensity changes,and noise interference.The experimental results show that the recognition accuracy under signal aliasing is more than 90%.The channel state can still be better distinguished under the condition of signal strength changes,and the prediction accuracy of time domain and correlation domain can still reach 88%.The channel state prediction effect is most affected by noise interference,and the overall recognition rate is over 80%.