Feature Extraction Of Infrasonic Signals Based On Multi-Scale Decomposition

Infrasound wave is a kind of sound wave whose frequency is below 20 Hz, which is outside the audible range of the human ear. Therefore, we can’t hear the infrasound wave in our daily life. Generally, earthquakes, volcanic eruptions, tsunamis, storms, thunderstorms and other natural disasters, as well as some human activities, such as explosions, guns fired, steamship sailing, cars, aircraft will produce infrasound wave, and some animals, such as elephants, tigers, crocodiles can emit infrasound too, even some organs of the human body can also produce infrasound wave. The infrasound wave has some characteristics, such as wide source, long transmission distance, low attenuation, strong penetration and so on. Moreover, it also has some dangers. Currently, infrasound is widely used in some fields such as weapons, medical treatment and earthquake prediction.In order to know more about the characteristics of infrasound wave and get more information from it, it is necessary to analysis infrasound wave by signal analysis techniques. With the development of the signal analysis techniques in the recent years, more and more analysis methods are put forward. The Fourier transform, Wavelet transform and Hilbert-Huang transform are some signal analysis methods which are commonly used, and people has used these methods in the analysis of infrasound wave. The wavelet transform and Hilbert-Huang transform are multi-scale signal analysis methods. In addition, this paper also applied a method which was recently proposed, namely the local characteristic scale decomposition algorithm, to the feature extraction of infrasound wave.The purpose of this paper is to study the effect of feature extraction of infrasound wave based on multi-scale signal decomposition methods. Firstly, the paper used the multi-scale decomposition algorithms to decompose the infrasound wave, and extracted the autocorrelation function, linear prediction coefficient and MEL frequency cepstrum coefficient of the infrasound wave to distinguish the infrasound from two different earthquakes. Secondly, the paper used the features to classify the two different kinds of infrasound from two earthquakes based on the Support Vector Machine algorithm. Finally, the paper analyzed the results of the experiments. And the experiment results have showed that the feature extraction based on the multi-scale decomposition algorithms is effective.