Seismic Event Detection Based On Smartphone Accelerometer

Seismic event detection is a key part of earthquake early warning system.The traditional detection technology determines the occurrence of earthquakes by analyzing the real-time data of the high-precision accelerometer fixed at the seismic station.In recent years,with the popularity of smart phones,seismic event detection based on mobile phone built-in accelerometer has become a new direction in the field of earthquake early warning research.Different from the traditional detection technology,seismic event detection based on mobile phone no longer focused on the extraction of seismic events from the strong noise data,but focus on mining the data model differences of human activity and seismic event.Therefore,the traditional seismic event detection method is not fully applicable in earthquake event detection based on mobile phone.Based on the characteristics of the typical human activity model,this paper designs a hierarchical seismic event detection framework mobile phone under fixed attitude of cellphones,and on this basis,through the data fusion between the accelerometer and the magnetometer,a seismic event detection method under non-fixed attitude of cellphones is proposed.The main work and contribution are as follows:Firstly,this paper presents a new seismic event simulation method based on seismic experience house to better reproduce the mobile phone acceleration data model under real earthquake scene.On this basis,the seismic event detection sample library is completed,which covers a variety of equipment,mobile phone gesture,mobile phone carrier,positive and negative sample type,and sensor types.Secondly,through the fusion of traditional threshold method and MyShake detection method,a framework of hierarchical seismic event detection algorithm is studied emphatically.At the lower level of the algorithm,the threshold algorithm extract the STA/LTA and SMC characteristics,and the typical human activity events with high seismic event discrimination are considered.At the higher level of the algorithm,random forest classification algorithm and time-frequency features are used,considering typical human events that are difficult to distinguish from earthquakes.The experiment results show that the hierarchical algorithm is effective in the detection of seismic events under the fixed attitude of mobile phone.Finally,on the basis of the hierarchical algorithm framework,the normalization of data coordinates of different mobile devices is realized by the integration of accelerometer and magnetometer data,and the geographically related data features are put forward to more effectively distinguish earthquake events and human daily activity event.On this basis,the seismic event detection strategy of mobile phone under non-fixed attitude is given:coordinate normalized data + geographic features+ random forest.The experimental results show that under this strategy,the positive and negative samples are more than 90%accurate,and the total feature dimension of the algorithm is greatly reduced.