Denoising Method of the ϕ -OTDR System Based on Wavelet-Subspace Joint Filtering

Considering the high cost of replacing high-performance components to address laser source frequency drift (LSFD) and noise from other optoelectronic devices in the phase-sensitive optical time domain reflectometer (
$\Phi $
-OTDR), a method based on joint denoising of wavelet transform and subspace techniques is proposed. The method aims to effectively remove low-frequency drift and random noise from signals, thereby improving the signal-to-noise ratio (SNR). First, the decomposition level and approximation component corresponding to low-frequency interference are determined through multilevel wavelet decomposition of the
$I/Q$
orthogonally demodulated signal. Subsequently, the approximation component is set to zero, followed by wavelet reconstruction, achieving the removal of low-frequency interference. Finally, random noise is suppressed using a time-domain-constrained subspace techniques. Footstep signals, heavy hammer striking the ground signals, shovel digging the ground signals, and speech signals were used to validate the effectiveness and applicability of this method. The comparison results with other methods indicates that this method exhibits outstanding performance in removing low-frequency interference and random noise. The SNRs of the four signals were improved by 31.55, 22.20, 18.71, and 26.49 dB, respectively. After calculations, it was found that the Pearson correlation coefficients (PCCs) between the waveforms before and after denoising for the four signals are all above 0.93. An analysis of the spectra before and after denoising indicates that the primary frequency components in the signals have not been lost. Additionally, experimental tests have shown that this method demonstrates outstanding robustness and stability.