Study On Weak Acoustic Signal Extraction Method Under Strong Background Environmental Noises

The radiated noise of vessel and submersible vehicles contains a lot of important information.These characteristic information can be used for the identification,tracking and parameter estimation of water surface and underwater targets,which can reflect their own operating conditions and target strength.Due to the strong interference effect of ocean background noise(waves,deep sea reverberation,etc.),it is very difficult to accurately extract the vessel's radiated noise.For the extraction of radiated noise with a large amount of information parameters in the strong ocean background noise,it belongs to the category of research on the extraction of weak acoustic signals under strong background noise.Therefore,obtaining the weak acoustic signal in the strong background noise is one of the important research issues for the identification of underwater targets such as ships and underwater vehicles.In this paper,based on the independent component analysis(ICA)algorithm in the blind source separation method,the time domain and frequency domain extraction methods of weak acoustic signals under strong background noise are studied.The numerical simulation method is used to realize the time domain separation algorithm of weak impulse acoustic signal under the background noise of single interference source and double interference source,and the weak acoustic signal is verified by experiment.Through numerical simulation and experimental methods,the effect of signal-to-noise ratio on the time-domain separation effect of weak pulsed acoustic signals is discussed.An analysis model of the frequency domain separation method for underwater weak acoustic signals in random ocean background noise is constructed,and the frequency domain separation of underwater weak acoustic signals is completed through numerical analysis and experiments.On the basis of the above research,based on the BP neural network algorithm to realize the pattern recognition of the time-domain separation results of the weak impulsive acoustic signals,a processing method for centralizing and scaling the separated signals is proposed.Aiming at the problem of weak acoustic signal separation under strong background noise,this paper achieves the extraction and recognition of weak acoustic signals under a certain signal-to-noise ratio,and has been practically applied in the field of ship and marine engineering.The specific research content and results are as follows:(1)Weak pulse acoustic signal separation method based on blind source separation method.The linear instantaneous mixed blind source separation model is analyzed.Based on the three basic assumptions of the blind source separation problem and the ICA method,several cost functions and optimization algorithms of ICA are introduced.Taking the realization of blind source separation under the condition of "total blindness" as the set goal,the fast fixed point method based on kurtosis is selected as the basic algorithm for the extraction of weak acoustic signals under strong background noise.Aiming at the degree of "weakness" of weak acoustic signals,the concept of peak-noise-ratio ratio is introduced,which is used for the subsequent research.(2)Research and experiment on time domain separation algorithm of weak signal under random strong background noise.Based on the PCA and ICA,using the fast fixed point iteration method,a time domain FastICA algorithm is established.Weak pulse acoustic signal is used as the signal source,and the separation methods under single background(random white noise)and double interference(random white noise + sine)background noise are studied respectively.The feasibility of the algorithm is verified.Weak pulse acoustic signal separation is achieved under the condition of low signal-to-noise ratio,and the effect of signal-to-noise ratio on the time-domain separation effect of weak pulse acoustic signal is analyzed numerically and experimentally.(3)Research and experiment on frequency domain separation algorithm of weak underwater acoustic signals in strong background noise.According to the characteristics of radiated noise generated by ships during navigation,the types of noise sources and spectrum analysis characteristics are introduced.Calculating the cost function and optimization method in the complex domain,the FastICA blind source separation method in the time domain is extended to frequency domain separation,and applied to the separation of propeller underwater radiated noise.The analysis of experimental test data verifies the feasibility of the algorithm.(4)Research on the pattern recognition algorithm based on the BP neural network weak acoustic signal time-domain separation results.Given the acoustic time-domain signal,the index suitable for the signal feature extraction is described.Taking the form factor,crest factor,pulse factor and margin factor as the time-domain feature index in this paper,the experimental test data is used to separate the signal for centralization and scaling After the feature index extraction,a 4-11-3 type BP neural network was constructed to perform pattern recognition of three types of acoustic signal sources,eliminating the interference of acoustic signal energy on the capture of time-domain features.The signal source type recognition and the pattern recognition of the weak sound signal under strong background environmental noise are realized.