Underwater variable frequency vibrating dipole detection of artificial lateral line based on the joint time-frequency analysis

Inspired by the lateral line system (LLS) of fish, an underwater pressure sensor array, aka, artificial lateral line system (ALLS) was proposed for underwater perception. The Fourier transform (FT) was the most widely used method to extract features from the ALLS signals. However, the FT lost efficacy when the vibrating condition of the dipole such as frequency varied over time. To overcome this problem, four joint time-frequency analysis (JTFA) methods were employed to transform the original pressure signal into the frequency domain and time domain. The short-time Fourier transform (STFT) and wavelet transform (WT) showed better performance in the dipole vibrating frequency evaluation. Later, the results of JTFA were input into a convolutional neural network (CNN) model to identify the motion pattern and position of the dipole. In this paper, different frequency rise/drop duration represented different motion patterns of variable frequency dipole. The position of the dipole indicated the coordinate of the dipole relative to the ALLS. The experiment result showed that the identification result of dipole motion pattern and position could reach 99.93 % and 2.134 mm respectively. Lastly, the ALLS pressure signals were discussed under the external flow and multiple dipole targets based on the tank experiment and computational fluid dynamics (CFD) simulation. The pressure drifting generated by the external flow was affected by both the flow velocity and the water depth variation. The distinction between JTFA results of different ALLS sensors exhibited correspondence with the state of multiple dipoles. The underwater feature extraction method of ALLS signal based on the JTFA proposed in this paper showed a promising way in the perception of the target in more challenging motion states.