The Application Research of Underwater Acoustic Source Numbers Estimation by Blind Separation Algorithm

The estimation of underwater acoustic source number is the basement of the blind separation algorithm of underwater acoustic signal. By using variance maximization of principal component analysis algorithm, multi-principal components can be extracted. Then three kinds of information theory index: AIC, MDL and Bayesian can be used to define the threshold of the principal component numbers that can be estimated the numbers of the source signal correctly. A series of simulation tests have been done which discussed the estimation accuracy of three kind rules under different signal noise rate. Moreover, the effect of the hydrophone numbers on the source number estimation accuracy also has been discussed. The results show that the algorithm has good noise robust and estimation accuracy in underwater acoustic signal source numbers estimation.