Multi-Source Separation Under Two “Blind” Conditions for Fiber-Optic Distributed Acoustic Sensor

Abstract

Significant progress has been made in single source recognition for fiber-optical distributed acoustic sensor (DAS). However, it is still challenging to detect and identify more than one unpredictable vibration sources when they are superimposed at the same fiber receiving point. Thus, in this paper it is proposed a blind multi-source separation method based on fast independent component analysis (FastICA), which utilizes the independency and non-Gaussianity of different sources. Firstly, two multi-source mixing mechanisms and separability of different sources received by DAS based on Φ-OTDR are discussed; to solve the two “blind” problems that the source number and the mixing mode are both unknown, a linear simultaneous mixing mode is assumed, and the source number is estimated by singular value decomposition to the observation matrix; then preprocessing of denoising and anti-mixing, and separation with FastICA by maximizing negative entropy are carried out to make the non-Gaussianity of the estimated signal achieve its maximum; finally, feasibility of the separation method is evaluated through several mixing cases including simulations with two to four field collected signals and a real field test with two sources superimposed on the buried fiber. Signal waves and the spectra, and three separation indicators, such as the Performance Index (PI), the signal correlation coefficients, and the signal mean square error (SMSE), are used to evaluate the performance of the method. As far as we know, it is the first time to realize the separation of an unknown number of the superimposed sources detected by DAS.