Earthquake event identification on smartphones based on principal component analysis and fully connected neural network

Currently, Earthquake Early Warning (EEW) requires expensive station construction and maintenance costs. With the widespread popularity of smartphones and the increasing accuracy of smartphones' built-in MEMS accelerometers, it has become possible to use smartphones as seismic monitoring instruments. To explore the feasibility of using smartphones for seismic event recognition, this study establishes a fully connected neural network model for distinguishing between daily human activities and seismic event signals recorded by smartphones. The dataset used in the experiments consists of seismic signals obtained by superimposing the internal noise of a cell phone on the ground shaking acceleration recorded by a station, and the acceleration data of daily human activities collected using a smartphone. In the data preprocessing stage, the original data were subjected to feature extraction, and then the extracted features were subjected to dimensionality reduction by principal component analysis, and the reduced features were used as inputs to the fully connected neural network. The results show that in the test set, the accuracy, recall, precision and F1 score for distinguishing seismic events from daily human activities respectively reached 95.24%, 93.62%, 97.80% and 96.08%, when using a 3-second P-wave hold; at the same time, the accuracy of recognizing seismic events with magnitude up to magnitude 7 and with epicentral distance of up to 200 km could reach more than 95%. Meanwhile, the accuracy of the model can reach over 95% for recognizing the daily activities of human beings when they ride on trains, automobiles, and subways, which are relatively smooth in movement. Based on the test results, it is inferred that it is feasible to utilize smartphones to recognize seismic events for EEW to a certain extent, and it can assist the existing seismic monitoring network, which also provides a new technical solution for upgrading the current EEW system in China.