Study On Blind Source Separation Of Underwater Acoustic Signals

Blind source separation(BSS)can recover the source signal only based on the observed signal but without knowing characteristics of source signal and channel parameters.It is of great significance and practical value in acoustic signal processing and image processing.Underwater acoustic signal and ambient sea noise are mutually statistically independent,the basic assumptions of blind source separation can be meeted,so blind source separation technique can be applied to the field of underwater acoustic signal processing.There are two uncertain problems when using blind source separation technology,one is uncertain amplitude,and the other is uncertain sequence.They have big impact on results,or even a separate failed result.In this paper,the two uncertain problems are analyzed and studied in depth.The main work and research contents are as follows:(1)Mixed signal model and classical blind source separation algorithm.Two kinds of mixed signal model are introduced,especially analyze the frequency-domain ideas of convolution mixture model.The separation performance and operation speed of the classical blind separation algorithm are compared through formula derivation and simulation experiment.Then sum up the differences and advantages and disadvantages of the two algorithms.(2)Method to solve the uncertain amplitude.In order to keep the expansion ratio of amplitude of separated signal and source signal consistent at each frequency point,the performance of two methods of minimum distortion and separation matrix normalization are studied.Simulation results show that the separation matrix normalization method is more stable in the full frequency range.As contrast,minimum distortion method can adjust reliable only in the local range.(3)Method to solve the uncertain sequence.In order to get the correct order of the separation signal at each frequency point,studies the effects of short term average amplitude and frequency range to interdependency.Experimental results show there are more correlation between short term average amplitude function and homologous signals than with heterologous signals.The farther the frequency is,the worse the reliability is.An improved ranking method is proposed in this paper by putting a frequency factor in the function.The simulation results show that the improved algorithm can adjust the order of each frequency point more effectively.(4)Blind source separation of underwater acoustic signals.The complex domain separation algorithm and the method of uncertainty are integrated into a complete blind separation algorithm in frequency domain,then analysis of characteristics of underwater acoustic signal,apply blind separation algorithm in frequency domain into the blind separation of underwater acoustic signals.Experimental results show that the frequency-domain blind source separation algorithm is effective.