High-precision sound source localization method based on successive approximation in an inhomogeneous sound velocity field

Abstract

In an inhomogeneous sound velocity field, the conventional time of flight (ToF) method cannot derive an exact distance directly. Therefore, a high-precision sound source localization method is required to address this difficult problem. In this study, we propose a novel high-accuracy successive approximation sound source localization method (SALM) in the inhomogeneous sound velocity field. The ToFs from a sound source to multiple receivers are used to constitute a real time vector. And the ToF from the assumed sound source to each receiver is considered a virtual time vector. It can be calculated according to the distribution of the sound velocity field, regardless of where the sound source is in the field. Then, modified vector and iterative formula can be constructed given the minimum difference between the roles of real and virtual time vectors. By using successive approximation method, the assumed sound source position approaches the actual position systematically until satisfies the localization accuracy requirements. Numerical results indicate that the SALM can considerably improve the localization accuracy. For example, in the 100 m × 100 m horizontal region, the ranging error of the SALM can be lower than 1 mm with more than 20 iterations when the velocity of sound is distributed unevenly within ±5 m/s. Moreover, increasing the number of receivers ensures a significant improvement in the approximate speed. And the localization accuracy can also be improved infinitely by increasing the number of iterations. Finally, the SALM is also applicable to sound source localization in a 3D space.