Detection of Underwater Moving Target in Low Signal-To-Noise Ratio Conditions Using Spectrogram Analysis

For underwater port security, the echo signals of underwater low-speed moving targets usually suffer from the low signal-to-noise ratio (SNR) which are induced by random noise or reverberation. Beamforming is conventional and very effective method to improve the SNR, but time-consuming and high computational cost with the number of receiving elements increasing. This paper presents a signal processing method based on frequency-wavenumber (or called FK) spectrogram analysis to improve the detection ability in low SNR environment. The method can simultaneously evaluate the velocity of moving targets, which is very useful when the doppler frequency is difficult to be measured because of slow moving speed of target or using such a doppler insensitive signal as source signal (e.g., linear frequency modulation signal). The spatial and temporal data are transformed into frequency and wavenumber domain through two-dimensional Fourier transform. The energy of moving targets in signals will concentrate on an inclined line distributed in the FK spectrum. The slope of the line represents the velocity of moving target. The procedure of beamforming is not necessary for the method, because the target signals can be easily separated from the echo signals by the FK spectrogram. The inclined line representing targets in the spectrogram is further extracted by Radon transform. Then the relationship between velocity and slope of the inclined line is built. We also give proof that the velocity calculated by the method actually is radial velocity of moving target through the first-order Taylor expansion. Furthermore, after the slope of line of moving target is calculated, a filtering operation is designed to separate the signals of target from the original and overlapped signals. Echo signals from the diver recorded in the environment of shallow sea are used to verify the method. The results show that the method can obtain an accurate estimation of velocity. The error of velocity is about 0.14%, compared with the velocity directly reading from the signals after beamforming. Additionally, strong noise is successfully suppressed after filtering. It is found that about 3dB gain is obtained after filtering.