Air Target Detection And Noise Reduction Algorithm For Passive Acoustic Research

Passive acoustic detector is an important part of the acoustic detection system. As the modern battlefield environment is getting more and more complicated, only using active detection technique is not sufficient. Because of the superior character of passive acoustic detector, it has attracted more and more people's attention.In this paper, we studied the base detection theory and the capability of single passive acoustic sensor. Although there are many well-worked signal detectors in theory, we select four different signal detectors for the object we should get and the type of data we have acquisited. They are energy detector, log sum (LS) detector, maximum power (MP) detector and harmonic set (HS) detector. The first two of them operate in the time domain, and others work in the frequency domain. We describe the four detectors in detail and also give the experimental results respectively. Their performances compared in terms of processing time and adaptability to environment change. According to the results, energy detector in the time domain and harmonic set detector in the frequency domain are recommended.Noise reduction of acoustic signal is another part of the paper. Under the influence of atmosphere, battlefield background acoustic signals may be highly unstable and non-gaussian. In order to increase signal-to-noise rate(SNR), we select recursive least-squares adaptive (RLS) filter and wavelet threshold de-noising method and we also study the two methods in detail. In the wavelet domain properties of wavelet coefficients of signal and noise may be different. Via threshold operation, noise can be restrained. (RLS) filter use prediction error to adjust its own impulse response until it becomes an optimized wiener filter. Based on simulations we use them to process the real acoustic signal. We also illustrate how to use the methods and an analysis of preferences.In order to improve detection distance, we use wavelet soft threshold de-noising method to process the data whose signal-to-noise rate is low before detection. We find a new way to calculate thresholds based on the pure background noise as well. In this way, the thresholds are more reasonable. The results are just as we expect, the SNR are improved and detection distances become longer.