The Design And Implementation Of Acoustic Detection And Positioning System

Target detection and recognition is a application technology,which synthesizes multiple-discipline methods,aiming at detecting the fixed or moving target by means of non-contact methods,and controling the object by the identification and positioning technology.At present,there are various methods of target detection and recognition,mainly including weapons and radar technology,infrared detection technology,the glimmer night-vision technology,laser technology,electro-optical countermeasure technology,etc.At the same time,the technology of sound detection is developing rapidly.It has been widely applied in civilian and military fields.By analyzing the collected sound signals,information about the sound,including category,model,distance and motion state can be captured.This thesis collects sound signals in a passive manner,and identifies the target by extracting features and classifications.If the recognition result is one of the targets,the positioning calculation is performed,and this information is sent to the central control machine to realize the target detection and identification,location tracking and pre-warning.The recognition of sound target is mainly composed of three parts: signal pre-processing,feature extraction and classification recognition.The positioning of target is a passive acoustic localization technique based on TDOA estimation,it take one of multi sensors as a benchmark and calculate the delay time between acoustic target to the benchmark and other sensors,so as to get the location information of the object,including distance,pitch angle,and azimuth angle.Therefore,this thesis is divided into two parts: acoustic target recognition and positioning.The identification of the target signal is relied on feature extraction and classification of thesound signal,which was collected by single channel sound transducer.The estimation of azimuth and elevation angle are realized by the four channel synchronous acoustic sensor.This thesis introduces methods of feature extraction,classification and recognition,time delay calculation,and the simulation results show that the Mel frequency cepstral coefficient method(MFCC)of feature extraction works better,BP(Back Propagation)neural network method of classification is more effective,and generalized cross-correlation method with maximum likelihood delay(Maximum Likelihood ML)window of time delay calculation bring a higher precision.In this thesis,the system hardware is designed for the acoustic detection system,and analysis as well as experimental tests were done on this basis.