Blind Separation Of Underwater Target Acoustic Scattering Signal

Underwater quiet small target detection and recognition is one of the most important problems in acoustic signal processing.The characteristics extraction of the ative sonar target echoes are the premise for the target detection,but it is affected by some critical problems,including the correctly analysis of the target acoustic scattering characteristics,settling the problem that there are the mutual interferences among target scattering components.Moreover,seafloor reverberation is the mainly interference for active sonar target echoes,especially for the target sinked to the bottom or buried,the buried layer absorbs the sound energy strongly,and the target echoes are weak,settling the problem that how to separate the target echoes in reverberation background and how to improve the signal to reverberation rate.In view of these two problems,reasearching the separation of the target scattering components in the free field to provide the stability characteristics,reasearching the target scattering components separation in reverberation background to improve the detection and recognition performance of the active sonar.Aiming at the problem that the underwater target scattering components are aliasing,a time and frequency domain blind source extraction method for the acoustic scattering components separation is presented,which is suitable for under-determined condition.Based on the target acoustic scattering physical mechanism,researching the time-frequency distribution characteristics of the target scattering components,constructing the appropriate structural element to filtering the cross-terms,the space time-frequency distribution matrices of the observed signals are obtained according to the target single auto-terms,and the corresponding acoustic scattering component is extracted by eigenvalue decomposition.And the extracted elastic scattering components are in good agreement with the surface wave theory calculation results.Simulation and experimental data processing results show that the algorithm can separate each of the target scattering components,and provide the stable and reliable characteristics for underwater target recognition.Aiming at the problem that the target echoes is seriously interfered by the seafloor reverberation,two kinds of blind source separation algorithms are presented.Researching the morphological characteristics of the target geometrical scattering components and the reverberation in Wigner-Ville distribution,a time-frequency blind separation algorithm combined with morphological filtering is presented.The expressions of the target geometric scattering components and the reverberation in the Wigner-Ville domain are deduced,andmorphological filter is used to remove the reverberation and the cross-terms between the geometrical scattering components in the Wigner-Ville distribution,obtaining an unmixed matrix in the time-frequency domain,so the target echoes are separated from the reverberation background.Based on the aggregation difference between the target echoes and the reverberation in the best fractional Fourier transform domain,a blind separation algorithm in fractional Fourier transform domain is presented.The objective function is constructed by the generalized correlation matrix in the best fractional Fourier domain,and the target acoustic scattering signals are obtained by approximate joint diagonalization.Compared with the time-frequency domain blind separation,it avoids the selection of the self-term time-frequency points and it can separate the target echoes in the lower signal-to-reverberation ratio(SRR)condition.In order to measure the separation performance of the algorithms,a SRR gain in the fractional domain is defined by the aggregation difference of the target echoes and reverberation,and it is used to quantify the separation effects of the blind separation algorithms.The echoes data from underwater suspension,bottom sinked and buried targets are analyzed.The results show that the target scattering signals can be separated in reverberation background.Compared with the performance of the two algorithms,for the underwater suspension and bottom sinked targets,blind separation algorithm combined with image morphology can obtain higher SRR gain.For the buried target,the reverberation interference is serious and the echoes are weak,blind separation algorithm in fractional Fourier domain can separate the target scattering echoes.For the active sonar target detection and recognition,the proposed two algorithms can effectively improve the SRR,and help to improve the performance of the underwater target detection.