Identification And Positioning Of Drone Using The Acoustic Microphone Array

The development of low-altitude aircraft technology has led to the widespread use of drones.However,its abuse also poses a serious threat to public safety.The drone has some features,such as low flying height,low flying speed,small volume and unobvious radar reflection.It has become difficult for low-altitude detection and prevention.This paper proposes a drone detection and direction tracking technology scheme based on acoustic signals for the countermeasures of drone.In the section of drone monitoring and identification,this thesis uses the MFCC technology for the acoustic signal feature extraction of small low-altitude vehicles.In addition,improving the mel filter used in MFCC technology referring to the characteristics that the low-altitude unmanned aerial vehicle has stable harmonic.By replacing the slope of conversion curve between linear spectrum and mel spectrum at such harmonics,the filter's sensitivity to the signal at that harmonic is increased.Then use the Gaussian mixture model to design the recognizer.Finally,the MFCC-GMM recognition method be used for the identification simulation of drone.The simulation results show that better recognition results can be obtained when using the improved MFCC method to detect and recognize the drone.In the section of localization,this thesis established a far-field array model of acoustic signals and focused on the SRP-PHAT positioning method.This method has better positioning performance because it combines the advantages of steered response power and time delay estimation.In view of the large amount of calculations in the global scanning of the SRP-PHAT method,this thesis improves the search strategy based on the dimensionality reduction search.A binary search strategy is proposed to reduce the calculations of the SRP-PHAT function by using the historical data of the upper layer scan.For the non-dimension reduction search strategy,the sparse distribution of sound sources in space is combined with SRC positioning method,using the delay difference of spatial point and microphone array to construct the localization model.Simulation results show that the binary search strategy can further reduce the amount of calculation required for positioning and the CSP-SRC method can get better positioning performance.In the section of direction tracking,this thesis analyzes the mechanism of the sound signal of drone which driven by propeller and the spectral distribution of the acoustic signal.LMS adaptive beamforming algorithm based on spectral envelope MMSE criterion is proposed for the characteristics that drone has steady sound signal spectrum envelope.The method calculates the array weight by seeking the minimum mean squared error between the reference spectral envelope and the output spectral envelope of the array and use the gradient descent method to update array weights.Simulation verified that the method can effectively track the drone target.