Seismic Signal Detection Method Based on Waveform Features

Abstract

The invention discloses a seismic signal detection method based on waveform characteristics, and relates to the field of seismic signal processing. The method comprises the following steps: firstly, selecting seismic signals and noise signals in historical events collected by an array as a data set, extracting an amplitude characteristic alpha, a ratio characteristic rho and a specific frequency band energy mean value characteristic gamma in each signal, normalizing, and normalizing an energy and characteristic lambda; dividing all seismic signals and noise signals into training samples and test samples; forming a corresponding matrix by the characteristic parameters of all seismic signals in the training samples, substituting the corresponding matrix into a Gaussian function, optimizing by using a gradient descent method to obtain an optimal hyper-parameter corresponding to each characteristic, and calculating a posterior mean value and a covariance of a Gaussian process of each characteristic to obtain four characteristic models; predicting the occurrence probability of a new event by using the verified feature model and a Bayesian thought, and judging whether the event is a seismic event or not according to the occurrence probability of the event. According to the invention, the correct detection rate is improved, and the applicability is stronger.

Description

technical field [0001] The invention relates to the field of seismic signal processing, in particular to a seismic signal detection method based on waveform features. Background technique [0002] In recent years, the emergence of station arrays has gradually replaced the method of using a single seismic station to collect data, and the amount of collected seismic data has become larger. The traditional method of using manual seismic signal analysis is inefficient, and the detection results are also susceptible to subjective judgments. The impact is no longer applicable to the era of rapid increase in data volume. [0003] Therefore, earthquake researchers have proposed a variety of earthquake detection algorithms, which can be roughly divided into two categories, one is a detection method based on a single feature, and the other is based on the overall feature or multiple features of the signal to complete the detection of seismic signals. For example, the long-short time ...

AI Technical Summary

Problems solved by technology

However, the long-short time window ratio method is not suitable for signals with low signal-to-noise ratio, and the F detection algorithm is more suitable for teleseismic signals with strong correlation

Images