Research On Gunshot Location Technology Based On Microphone Arrays

In the battlefield and terrorist attacks,the sniper’s cold shot is often a powerful tool to take lives and put people in a panic.The anti-sniper detection and positioning system can quickly find the sniper’s location and counterattack,and once became a research hotspot in various countries in Europe and the United States.While China’s research on anti-sniper systems is still in its infancy,in order to improve this status quo,this thesis investigates the microphone array-based gunshot localization technology,including gunshot detection,identification and localization techniques.The main work is as follows:In the detection of gunfire,due to the unfamiliarity of gunfire signals,the existing research basically draws on the methods for processing speech signals,which are not all suitable for gunfire.To address this problem,this thesis starts from the formation principle and acoustic characteristics of gunfire,carries out suitable pre-processing of gunfire signals,and realizes effective detection of gunfire signals using endpoint detection technology to lay the foundation for the recognition and localization of gunfire.In the recognition of gunshots,the use of threshold values in the time-frequency domain to discriminate gunshots has a high false alarm rate,poor targeting,poor portability of the algorithm and is not stable enough.The recognition method using feature vector decomposition is too computationally intensive and has low real-time performance.The features extracted by the gunshot recognition method using deep learning are not well targeted to gunshots.To address these problems,this thesis extracts features such as feature vector MFCC commonly used for speech signal recognition,speech spectrogram containing time-frequency domain information of gunshot signal,short-time transient rate,and short-time energy for comparison,respectively.Through the fusion recognition of the one-dimensionally arranged Meier speech spectrogram features,short-time transient zero rate and short-time energy features,the gunshot signal can be well distinguished from other similar high pulse signals,and the false alarm rate of gunshot recognition is significantly reduced.In the localization of gunfire,most of the traditional gunfire localization techniques use rifled wave localization,ignoring the information carried by Mach waves.The fusion algorithm of spatial gridding search is computationally complicated when fusing the results of rifled wave multi-base array localization.In the localization method based on Mach waves,the azimuth angle can only be solved by Mach angle information,which is very limited.To address these problems,in the rifled wave localization,the upper and lower symmetric orthotetrahedral array elements were selected to reduce the angular error,and the sniper position was solved by establishing the hyperbolic localization equation based on the time difference estimation of the gunshot signals received by different array elements.The Geometric Dilution Precision(GDOP)and the Cramer-Rao Lower Bound(CRLB)of the algorithm are derived,and the feasibility of the algorithm is demonstrated from theoretical analysis to simulation analysis.The proposed weighted algorithm fuses the boresight wave localization results of multiple microphone arrays,which improves the localization accuracy and reduces the complexity of traditional localization fusion algorithms.In Mach-wave localization,multiple time-delay estimates of Mach waves are proposed for localization based on the short-time motion model of projectile flight,and finally,after processing the projectile trajectory by Kalman filter,the sniper’s position is derived using least-squares estimation based on the flight trajectory,and the CRLB lower bound of the algorithm is derived.