Learning-based blind passive weak signal detection using a particle motion vector sensor in unknown time-spreading distortion underwater channels

Signal detection in underwater channels is a challenging task, particularly when dealing with time-spreading distortion (TSD). The challenge becomes more complicated when aiming at blind passive signal detection, where both the signal and the TSD channel are unknown. In this research, we utilize a dictionary learning (DL) approach to perform blind passive signal detection, leveraging the sparsity of underwater channels impulse responses. We conducted underwater experiments to evaluate the performance of the DL-based detection method and to compare the results with conventional detection approaches. The data were collected using GeoSpectrum's M20-040 particle motion sensor along with its analog box and extension cable. This vector sensor comprises three orthogonal accelerometer dipole sensors and one acoustic pressure omnidirectional sensor. The high detection probabilities obtained by our learning-based method using one single compact sensor exhibit its usefulness in detection scenarios where the signal and the channel response are not known in advance.