Research On Acoustic Target Recognition Technology In Field Environment

In the field environment of frontier defense area,the detection and identification of invading target are accomplished through analyzing and processing the sound signal of invading target,so as to realize the intelligent real-time monitoring of frontier area.The traditional target detection and recognition method USES multiple sensors for combined detection.Among them,the sound sensor is easily interfered by noise and other sound sources,which leads to the decrease of the accuracy of sound target recognition.Aiming at the above problems,this paper designs a vehicle and human identification scheme in the field environment,and optimizes and improves the endpoint detection algorithm.This paper focuses on the acoustic target recognition technology based on sound sensor and discusses it from the following four parts:(1)In the field environment,the acoustic target signal collected will inevitably be interfered by background noise,making the target signal lose its original characteristics.Therefore,it is necessary to reduce the noise of the collected signals in order to improve the signal-to-noise ratio of target signals.Since the background noise in the field environment is complex and changeable,the adaptive noise cancellation algorithm based on RLS and LMS is adopted for noise reduction processing,and the simulation experiment results are compared and analyzed.Finally,the RLS algorithm with better convergence speed and convergence accuracy is selected to complete the signal noise reduction processing.(2)After noise reduction processing of the collected sound signals,endpoint detection of the acoustic target signals appears.When there are intrusive acoustic targets,the improved endpoint detection algorithm based on short-time energy is used to detect them.Through a long frame coarse detection and a short frame fine detection,the starting and ending points of the target signal are precisely located,and then the invalid background noise segment is eliminated.In addition,a threshold decision scheme based on the short-time average range is designed to pre-determine whether there is an intrusion target after the endpoint detection is completed.(3)The basic principle and steps of extracting time-frequency domain characteristic parameters were briefly described,and the applicability of different characteristic parameters in acoustic target recognition was compared.Through the analysis of different characteristic parameters and the comparison of subsequent experimental results,the MEL frequency cepstral coefficient is selected as the characteristic parameter of vehicle and human identification.(4)Select an appropriate classifier for the acoustic target recognition system to complete human and vehicle target recognition in the field environment.In this paper,based on the characteristics of moving target's sound signal and that of talking human,and taking the MFCC of vehicle's sound signal as the characteristic,a classifier based on Gaussian mixture model(GMM)is adopted to realize vehicle target recognition.Finally,through the comprehensive analysis of experimental results and field test results in the field environment,it is proved that the recognition scheme designed in this paper can complete the identification of human and vehicle in the field environment,and both the recognition rate and the recognition speed can meet the actual requirements.