Research On Technology Of Gunshot Localization Based On Compressed Sensing Theory

In modern wars,sniper rifle has good concealment and other good characteristics,terrorists use these characteristics of sniper rifle to pose a great threat to the other army.With the gradual development of sound source localization technology,it is an effective method to determine the sniper position through sound source localization technology.The conventional sound source localization methods are roughly divided into three types,which include sound sensor positioning method based on steerable beamforming,subspace-based sound source localization method and sound source localization method based on time difference of sound.Among them,the existing gun sound localization methods are basically based on the time-differentiated positioning method,but this method has the disadvantages of poor anti-noise performance and low recovery accuracy,which is difficult to meet the requirements of modern anti-sniper.With the rapid development of compressed sensing?CS?theory in signal processing,the DOA estimation based on compressed sensing theory has the characteristics of small amount of collected data,good anti-noise performance and high recovery accuracy.Therefore,it is proposed to spatial DOA estimation of gun sound sources based on CS theory.However,there is no corresponding theoretical research on the application of CS theory in spatial sound source distance estimation.Combining actual anti-terrorism requirements,the importance of DOA estimation is much greater than the estimation of spatial distance.Therefore,it is mainly determined to the position of gun sound source from the combination of DOA estimation based on compressed sensing theory and spatial distance measured based on delay difference method.Specific research work includes:1.At present,DOA estimation based on compressed sensing theory is mainly based on the study of 2D linear array models.There is no mature theory and research on the application of DOA estimation in 3D space.The ability to obtain sparse representation of the signal is the key to the application of compressed sensing theory for spatial DOA estimation.Firstly,several kinds of spatial source positioning array models were analyzed.Based on the combination of mathematical derivation and compressive sensing theory of each model,the L-array model is selected as the spatial acoustic source localization array model.2.The array manifold matrix is deduced by the geometric relationship of the array sensors in the L array.Based on the compressed sensing theory,the array manifold matrix is transformed into a sparse matrix.The L-array can be regarded as consisting of two mutually perpendicular linear arrays,so two sparse matrices are obtained.In both sparse matrices,the azimuth and elevation angle information are included.Two-dimensional linear arrays have sparse matrices with two grid divisions.There are four ways of dividing into three-dimensional spatial grids.The four partitions recovery DOA estimation performance is verified by whether it satisfies the RIP property well,but the RIP condition is theoretically verified as an NP-hard problem.Therefore,it is verified to the column vector of sparse matrix according to the RIP conditional equivalent transformation proposed by scholars.Orthogonality verifies the orthogonality of the four division patterns.3.A weightedl₁algorithm based on compressed sensing sparse reconstruction is proposed.This algorithm can overcome the influence of coherent sources at low SNR and better suppresses the false peaks that appear when reconstructing the signal based on the unimproved pre-₁lalgorithm.Secondly,the performance of the algorithm is analyzed from the complexity and stability of the algorithm.Finally,this algorithm is applied to the signal reconstruction of the four spatial partition models,and it is determined that the space partition model with equal sine is the best.4.The estimation of the spatial distance of sound source is an important research direction of spatial positioning.At present,the mature theory is based on the time delay method to measure the time delay difference of the received signal of the sensor and combine the geometry relationship of the element to estimate the distance.The more mature method of delay measurement is based on the autocorrelation and cross-correlation measurement delay method.The cross-correlation is less affected by the source and noise correlation,and the anti-noise performance is relatively good.However,there is still a large error in the case of low signal-to-noise ratio.it is used to an improved PHAT-weighted cross-correlation delay method to measure the delay.In improving PHAT-weighted functions,using the natural noise power spectrum databasewithout gunfire environment,the noise power spectrum function in the corresponding scene model is removed from the weighting function to obtain an improved weighting factor combined with the actual scene model.Simulation experiments show that the improved PHAT weighted cross-correlation delay method has good performance under low SNR conditions...