The Research On Infrasound Of Disaster Based On Signal Feature Extraction And Clustering Algorithm

During the development of disasters,such as earthquakes,tsunami,volcanic eruptions,infrasound waves with unique characteristics are generated.At the same time,clustering algorithm is one of the key application technologies in data mining algorithms.It can be used to find some clusters in disordered data sets.By means of analyzing the characteristics of infrasound waves generated by different types of geological disasters in the atmosphere and extracting the characteristics of the collected geological disaster infrasound signals,it is able to achieve the best feature of geological disaster infrasound signal,which make the same type of geological disaster infrasound signal similar,and different types of geological disaster infrasound signals obvious distinctive.To find apparent features and distinguish the different types of geological disasters is definitely significative to improving the accuracy of monitoring geological hazards.According to the infrasound signal clustering system,the collection of digital infrasound signals of geological disaster is the basis of this study,and infrasound feature extraction technology and clustering algorithm are the focus of this study,and evaluating clustering methods is the key to reaching a reasonable conclusion.Through Coordinated Universal Time,the geological disasters infrasound signal can be accurately located.Digital infrasonic signal data is provided by the Comprehensive Nuclear-Test-Ban Treaty Organization and obtained from the Beijing National Data Center of the Ban-Terrain Test.As the representative methods of time-frequency analysis in digital signal processing,wavelet transform and Hilbert-Huang transform are used in this study of the infrasonic signal feature extraction techniques.Wavelet transform is selected to use the energy spectrum of digital infrasound infrasound signal as one of the infrasound signal feature vectors.The Hilbert-Huang transform is used to extract the multiple eigenmode functions of the infrasound signal,and the significant frequency in the marginal spectrum of the intrinsic modal function signal is taken as another feature vector of the infrasound signal of the terrestrial disaster.Bycomparing the accuracy,running time,and CH index of the partition-based clustering algorithm,hierarchical-based clustering algorithm,the density-based clustering algorithm,the model-based clustering algorithm,the feasibility and effectiveness of the application of infrasound signal clustering system in distinguishing natural disasters can be verified.In this study,three infrasounds signal,earthquakes,tsunamis and volcanoes,are tread as subjects.The features of infrasound and clustering algorithms are used as variables.It can be summarized that AGNES clustering algorithm have the highest accuracy reaching 0.93,when the significant frequencies of IMF components are regarded as infrasound features,Manhattan metrics and Average Linkage were used as parameter.The result of study proves that the infrasound signal clustering system can be used to distinguish the types of disasters.It is expected that the results of this study have a certain reference effect on the practical application of infrasound signals.