An adaptive model for the location of low-frequency sound sources in shallow sea environments

Abstract

At present, mathematical models applicable to the localization of shallow sea low-frequency sound source targets are usually complicated, and some environmental parameters need to be set in advance. This poses a great challenge to the scope of the application of this technology. To solve this problem, we propose an adaptive sound source localization model (ASSL) based on data driving, which only needs to input the vector hydrophone signal to output the two-dimensional coordinates of the sound source in real time. First, the four-channel sound source signal received by the single vector hydrophone is fused into a single-channel sound intensity signal and dynamically divided into suitable lengths based on Shannon entropy; then, the built deep bidirectional long short-term memory (DBiLSTM) network is used for a priori training; finally, the results are cross-validated, and the location of the target sound source is output. Experimental results show that the model can quickly estimate the position of the low-frequency sound source in the shallow sea environment using only a 1 s signal collected by a single vector hydrophone. The error between the estimated position and the actual position is approximately 35 m. Even with large samples, the method also showed great potential for application.