Research On Network Detection And Recognition Technology For Target Tracking

With the rapid development of science and technology,wireless sensor networks play an important role in many fields such as military,civilian and so on,and bring great convenience to people's life.Target tracking is one of the important applications of wireless sensor networks.It is extremely important to detect and recognize target before tracking,which mainly include getting the real-time information of target category and location accurately.At present,the active sensors such as radar have better performance of target recognition and location,but its concealment is poor and it is easily to be intercepted.However,the acoustic sensor has strong concealment and can work in passive mode.Furthermore,the sound signal is not blocked by obstacles,so that the acoustic sensor has important applications in the field of target detection and recognition.Therefore,this paper focuses on the technology of acoustic target signal recognition and location,and suggests counter measures against deficiencies in existing methods.The main research contents include the following four parts:(1)Based on the research of the technology for passive target detection and location in recent years,the significance of the research on the recognition and location of the acoustic target is clarified.According to the research status of acoustic target recognition and positioning technology,the key technology is analyzed and the technical problems are pointed out;(2)The signal needs to be preprocessed such as noise reduction before identifying and positioning the target.This paper analyzes the basic theory of different wavelet de-noise methods,and different wavelet threshold setting methods are simulated and verified.The wavelet soft threshold method is used to de-noise the measured target signal;(3)In stage of identification,mainly divided into two parts:(1)feature extraction,a method of feature extraction based on combined features and LDA is designed to extract features effectively.The characteristics are analyzed,and the time domain features(short-term energy,short-time average amplitude and zero-crossing rate)and frequency domain characteristics(energy of each band extracted by the wavelet packet)of the signal are extracted to combine the time domain and frequency domain features.The method of linear dimension analysis is used to reduce the dimension of the extracted feature vectors for the high-dimensional characteristics of the combined features;(2)classification and recognition,two classification algorithms consist of support vector machine and K nearest neighbor are used for target recognition.Different feature extraction methods for the acoustic signal of two types of target consist of cars and buses are compared by simulation.The results verify that the proposed method has higher target recognition rate;(4)In stage of source localization,the algorithm of location based on time difference of arrival is adopted.The algorithm divided into two parts:(1)time delay estimation,this paper analyzes the time delay estimation method of weighted generalized cross correlation,aiming at its shortcomings,a new method based on the combination of phase difference and maximum likelihood is proposed to improve the accuracy of time delay estimation in the case of relatively low SNR;(2)position estimation,the CHAN algorithm is used for estimating the target position based on the time delay estimation.The proposed time delay estimation method is verified by the simulation,the result verifies the effectiveness of the proposed time delay estimation method and with high positioning accuracy.