Location Method Of Trapped Personnel After Earthquake Based On Microphone Array

In recent years,earthquake disasters occur frequently.If buildings collapse after earthquakes,ruins will be formed.The trapped people buried in the collapse of buildings are one of the main objects of emergency rescue after the earthquake.A large number of voids often exist in the collapsed buildings and ruins,and trapped people can send out sound signals for help.However,the structure and environment of the ruins are complex,and the location of the trapped people can not be accurately judged,which makes it difficult for timely and accurate rescue.At present,the microlife detector in China can detect the voice of the trapped people,but it can not accurately locate the trapped people.The location of trapped people after the earthquake has become one of the hot issues in emergency rescue research.Aiming at the problem of inaccurate location of trapped people after earthquake,this paper combines acoustic source location with neural network technology,establishes a geometric model of six-element microphone array in ruins environment,calculates the difference of time delay based on the geometric model as input,and takes the location of target sound source as output,establishes a prediction model of neural network location,and simulates the trapped people through the simulation model.The prediction of the location of the people in distress can improve the accuracy of location and narrow the scope of post-earthquake rescue.On the basis of studying the principle of microphone sound source localization,this paper compares three kinds of sound source localization technology,and chooses the localization technology with less calculation and better real-time performance in the post-earthquake ruins environment.In this paper,by building two-dimensional and three-dimensional spatial microphone array geometric models in the ruins,the shortcomings of two-dimensional plane array model in spatial positioning are analyzed.By comparing the positioning performance of three three three three-dimensional spatial microphone array geometries,the geometric structure of microphone array with the best positioning performance is selected.In this paper,the basic theory of RBF neural network is studied,and the learning algorithm of RBF neural network is analyzed.It is confirmed that RBF neural network has good fitting ability.Based on the characteristics of radial basis function(RBF)neural network applied to multi-input multi-output(MIMO)system and nonlinear system,the simulation model of neural network location prediction is established.According to the geometric relationship of the six-element microphone array structure,the time delay difference is used as input and the given coordinate of sound source is used as output to determine the input and output layers of the simulation model.According to the specific situation of building collapse and people buried in the earthquake scene,the data samples of the simulation model are extracted,and the samples are pretreated,so as to facilitate the learning,training and prediction of the simulation model.In this paper,the positioning simulation model is debugged on the platform of matlab,and the conventional parameters are given.By training and predicting the simulation model,the predicted value can greatly improve the accuracy of positioning,narrow the scope of rescue and improve the efficiency of rescue,which verifies the correctness of the positioning simulation model.At the same time,by constructing BP and GRNN neural network simulation models,and comparing with RBF neural network location simulation model,the paper proves that the RBF neural network acoustic source location simulation model has better positioning performance.Considering the arbitrary position of the microphone array in the post-earthquake ruins,the influence of the array geometry on the location prediction effect of the simulation model is also studied.By establishing three regular geometric structures of three kinds of three-dimensional spatial microphone,the effects of array structures of these three rules on azimuth angle,pitch angle and distance positioning accuracy are simulated and analyzed.In the research process of this paper,the GUIDE simulation interface is established under the GUIDE graphical user development environment of Matlab software,and the simulation platform of trapped person location prediction based on neural network is established.In the environment of earthquake rescue,search and rescue personnel can realize human-computer interaction through simulation platform.