Research On Ship-radiated Noise Detection Based On Time-frequency Analysis

Ship-radiated noise is broadband noise source. It is of great significance for targetdetection and recognition which has been a great problem. The research of ship-radiatednoise becomes a hot topic among domestic and foreign scholars. This paper studiesship-radiated noise detection technology based on time-frequency analysis by processingsimulation signals and experimental datas to verify the feasibility and effectiveness. The maincontent of this paper include the following aspects:1、The paper introduces three kinds of filtering method in wavelet domain. This partcompares the effect of filtering by simulation signals and analyzes the advantages anddisadvantages among the three methods.2、The paper discusses the basic principles of the Hilbert-Huang Transform and defectsof the algorithm. It also gives K-L divergence to identify false IMF components bysimulation signals to verify the effectiveness. Because of EMD having adaptability andfilter properties, it can extract the signal trend items.3、The paper analyzes the basic theorys, basic algorithms and filtering principles ofsparse decomposition. It also introduces several kinds of typical dictionary of atoms.MapFOCUSS algorithm is utilized to eliminate transient pulse interference. The effectivenessof this algorithm is verified by simulation signals and experimental datas. Because traditionalsampling needs more datas, the paper studies the basic theory of compressed sensing.4、The paper uses cosine dictionary of atoms and time-frequency dictionary of atoms todetect the target signal frequency through simulation signals and experimental datas, and itobtains good results. The detection effect of sparse decomposition based on time-frequencydictionary of atoms is good, but the calculation speed of the algorithm is too slow. For thisreason, the introduction of particle swarm algorithm improves the calculation speed greatly.Finally, in the case of ensuring the detection capability, compressed sensing combines withsparse decomposition to detect the target signal for larger datas. It can reduce the dimensionsof datas and improve the calculation speed at the same time.