Low-noise phase demodulation algorithm for fiber optic interferometric sensors

This is the authors' abstract. We don't add key points for this paper.

A low-noise phase demodulation algorithm based on 3 × 3 coupler demodulation is proposed for interferometric fiber-optic sensing systems. This algorithm uses the covariance matrix of the three-channel first-order difference vectors to construct a statistical weighting matrix, then combines it with the covariance matrix of the original observations to obtain a two-dimensional weighted projection through generalized eigenvalue decomposition, and finally integrates the ATAN algorithm with phase unwrapping technology to recover the measured phase. Experimental results show that this algorithm reduces the demodulated noise floor under both single-frequency and dual-frequency excitations, with a maximum reduction of approximately 14 dB to a level of 1.8 μrad/√Hz at 30 kHz. For speech-signal demodulation, the spectrogram exhibits reduced background noise, and the short-time objective intelligibility is improved from 0.77 to 0.97. The algorithm can still support stable ellipse parameter extraction and maintain a linearity of 99.99% when the excitation amplitude is as low as π/8 rad.