On the Feature Phase Space in the Seismic Pattern Recognition of Underground Nuclear Explosions

Abstract

We present the origin of the equivalence of feature space and phase space based on the determination of the minimum of uncorrelated features and the choice of training samples in the seismic pattern recognition of underground nuclear explosions. The validity of the equivalence of feature space and phase space is tested by numerical experiments. Experimental results indicate that the attractor analysis approach of feature phase space is an efficient way for pattern recognition. It determines the lower limit of uncorrelated features in number, has great immunity to noise, and is very suitable to analyze practical samples with noise and disturb. At the end, the idea of further research is presented.