Research On Time-varying Feature Extraction Of Underwater Diver Radiated Sound Based On Vector Signal Processing

Feature extraction of underwater small target is a prerequisite for its detection and recognition.Underwater divers are one of the primary threats to anti-terrorism and underwater security in recent years.It has very different acoustic signal characteristics from the tradition underwater targets such as ships or torpedo.Its signal is lo-w power,low SNR,usually has the characteristics of regular changes over time.How to effectively extract diver features from background noise is the key points and difficulties in the research.In this paper,in the light of passive detection of divers,contrasting divers with open-SCUBA and divers with ADS,a brief description of their different breathing principle and underwater acoustic radiation physical source is given.Compared to the conventional low-pass filtering demodulation and a filter detector envelopment analysis,paper focuses on the theoretical basis and mathematical principles of Multiband Matched Filter algorithm.The vector signal processing technology is introduced into MF algorithm to optimize the performance.Using the simulation signal analysis the SMF and NMF output performance of suppressing the background noise and strong interference.A Gamma value calculated from the spectral envelope of MF output is given as detection reference.Then the study analyzed the trends of Gamma value in gradient SNR case,verified the advantages and feasibility in MF based on vector signal processing.Using the vector hydrophones designed and completed the diver test in the experimental pool.Obtained diver radiated acoustic vector signals in different situations.Combined the time,frequency,and time-frequency domain,using the vector analysis method to study diver acoustic radiation characteristics.For divers with open-SCUBA,gives a stable phase and enter water phase analysis;for divers with ADS,gives analysis of air-venting sound and transient pulse signal.Comprehensive compared of the signal differences of these two kinds of divers,effective conclusion is given.First using general envelopment analysis to extract the diver radiated acoustic energy envelope,spectral envelope is given and the performance analysed.The MF algorithm based on vector signal processing was applied to test pool data analysis,both SMF and NMF output performance in the actual noise environment is tested.A GMF output is proposed according to actual needs to boost output stability.A study about two main parameters of MF algorithm,the time window width and number of subbands,is given to find the best value of performance for diver signal analysis.Use GMF obtained regular change energy envelope of diver radiated acoustic signal,got a clear,stable envelope spectrum.Extracted the breathing frequency of diver with open-scuba and air-venting frequency of diver with ADS,verified the accuracy and validity of the MF algorithm.Calculated the Gamma value from spectral envelopes based on different methods,identified GMF output has the biggest advantage in diver radiated acoustic energy envelope extraction.The conclusion has practical reference value for further research about diver detection,localization and tracking technology.