SSL system based on microphone array [1] relies only on receiving acoustic signals from the target to determine its distance and orientation, which has unique advantages such as strong concealment, high security, and not easy to be detected, and thus is a hot spot of the current research on SSL [2]. In most cases, people choose to use an electrical microphone to realize the SSL function [3]. However, compared with electrical sensors, fiber optic sensors [4] have the advantages of flexible structure, easy long-distance transmission, high sensitivity, and strong anti-interference ability. In recent years, the use of fiber optic sensors to achieve SSL in complex situations and serious electromagnetic interference environments has received increasing attention in research [[5], [6], [7]]. Among them, the (EFPI) sensors [8] has gained wider attention on this research field due to their simple structure, smaller size, and higher performance.

There are three main methods for SSL based on microphone arrays: Controllable beamforming technology [[9], [10], [11]] takes the output power of the array beam as the basis for modeling and finds the maximum point of the region of controllable output power as the location of the sound source by phase weighting and transforming. This method can be used for the localization of multiple sound sources, but the computation is large. The high-resolution spectral estimation technique [12] calculates the azimuth of the signal based on the correlation matrix between the signals received by each microphone to determine the location of the sound source. The method can achieve high accuracy in theory, but the algorithm complexity is high. The SSL method based on TDOA [13] first estimates the time difference between the arrival of the sound signal to each microphone in the array, calculates the approximate distance difference from the sound source to each microphone by this time difference, and finally estimates the position of the sound source by combining with the geometric localization algorithm. Due to the high accuracy and fast calculation speed, such methods are widely used in SSL. In this paper, to build a real-time localization system, the TDOA method is chosen to calculate the location of the sound source.

Among the demodulation algorithms for Fabry-Perot fiber-optic sensor [14], three-wavelength adaptive intensity demodulation [15], as an intensity demodulation algorithm, greatly improves the dynamic range of demodulation while maintaining the demodulation speed and is suitable for real-time positioning systems. However, it is not directly applicable to TDOA localization techniques because temperature variations as well as humidity variations change its demodulation region [16], resulting in a possible flip of the final signal. Based on the above, this paper analyzes the reasons that lead to signal flipping and proposes an adaptive correction method to make it applicable to SSL.

In this study, we use four EFPI acoustic sensors to build a 3D microphone array [17] and use the proposed adaptive method to modify the three-wavelength adaptive intensity demodulation method to obtain flip-free signals, and then calculate the position of the sound by TDOA technique. The system does not require complex instrumentation and algorithms, thus obtaining lower complexity and faster computation rates, enabling real-time localization of sound.