Research And Realization Of Battlefield Acoustic Target Real-time Recognition And Location Technology Based On DSP

The accurate recognition and location of battlefield targets,such as tanks,wheeled chariots,armed helicopters and fighters,are key means to win the control of the battlefield,and are of great value in modern military affairs.The target recognition and location based on passive acoustic detection have the advantages of unlimited visibility,good privacy and strong anti-interference ability,so it has become the research hotspot of the military powers.Due to the complexity of the battlefield environment and the real-time requirement of the system,how to realize the real-time high precision location and high accuracy recognition of the battlefield acoustic targets have become a challenging research topic.This topic mainly studies the algorithm of battlefield acoustic target recognition and location,and based on DSP,the software and hardware design of real-time recognition and location system for battlefield acoustic targets is completed.The feature extraction of acoustic signals is the key to the acoustic recognition.The composition of the acoustic signal in the battlefield is very complex,which contains both the random and the quasi-periodic signals.Based on the in-depth analysis of the generating mechanism of several typical battlefield acoustic target signals,a multi-feature extraction algorithm was proposed by combining wavelet packet analysis with discrete spectrum analysis.The algorithm utilizes the excellent time-frequency local analysis ability of wavelet packet transform to extract the energy distribution of acoustic signals on the non-uniform frequency bands,and extracts the fundamental frequency and harmonic characteristics describing the quasi periodicity of acoustic signals by discrete spectrum analysis,so as to reflect the characteristics of the acoustic targets more comprehensively.In order to improve the signal to noise ratio,the pretreatment method of wavelet packet de-noising is used to reduce the influence of environmental noise on acoustic signals before feature extraction.The BP neural network is used to classify the feature parameters,the simulation results showed that the accuracy of the multi-feature extraction algorithm applied to acoustic recognition is improved obviously,compared with that of traditional single analysis domain feature extraction algorithm.According to the battlefield acoustic target location,this paper introduced the advantages and disadvantages of several common acoustic source localization algorithms,and the acoustic source localization algorithm based on time delay estimation is determined.Time delay estimation is the key technique of acoustic source localization,the generalized cross correlation algorithm can suppress the influence of noise on time delay estimation,the time delay estimation algorithm based on generalized cross correlation was mainly studied.Because of the accuracy of time delay estimation is affected by the sampling frequency of the signal,the parabola interpolation method is used to improve the accuracy of time delay estimation.Aiming at the individual outliers in the location results,a post-processing of the median filtering and the least-squares-fitting filtering was performed on the location results.Simulation experiments of acoustic source location algorithm and post processing algorithm were carried out.Compared with the actual position of the target calculated by the GPS timing,the accuracy of location was analyzed,and the effectiveness of post processing algorithm was verified.In order to recognize and locate battlefield acoustic targets in real time,the software and hardware system of acoustic target recognition and location was designed by using TMS320VC5510 as the core processor.The hardware system includes analog signal processing circuit,DSP digital signal processing circuit,DSP peripheral auxiliary circuit and power circuit,etc.Aiming at the problem that the 16 bits fixed-point DSP is limited in dealing with the decimal fraction,the Q value calibration scheme of each parameter is given.The method of floating-point operation to fixed-point implementation and the overflow processing method of fixed point operation are discussed.In view of the high complexity of the algorithm,which affects the real-time performance of the system,the program of some modules was optimized by mixed programming and other methods,thus the speed of its operation in DSP is improved.The results showed that the system satisfied the performance requirements of recognition rate,positioning precision and operation time,and the satisfactory results were obtained.