Characterization of DWT as denoising method for φ-OTDR signal
View/ Open
Date
2021-12Author
M. S., Yusri
B., Faisal
A., Ismail
N. L., Saleh
M. F., Ismail
N. D., Nordin
A. H., Sulaiman
F., Abdullah
M. Z., Jamaludin
Metadata
Show full item recordAbstract
DAS system based on φ-OTDR technique suffers from random noises that affect the signalto-
noise-ratio of the extracted signals. This results in high false alarm rate, reducing the
capabilities of the systems to detect vibration signals. This paper presented a thorough
analysis of a denoising method using discrete wavelet function (DWT). We implemented
and compared different mother wavelets such as Symlet 4, Haar, Daubechies 4 (Db4),
Biorthogonal 4.4 (Bior4.4), Coiflets 3 (Coif3), Discrete approximation of Meyer wavelet
(dmey), Fejér-Korovkin filters 8 (fk8) and Reverse Biorthogonal 6.8 (rbio6.8), using
multiple levels of decomposition. Four denoising thresholds, Empirical Bayes, Universal
Threshold, Stein's Unbiased Risk Estimation (SURE), and Minimax Estimation (Minimax)
were characterized using soft threshold rule. From the results obtained, the combination of
the Daubechies 4 wavelet function, level 3 decomposition, SURE denoising threshold with
soft threshold rule produces the best denoising performance on the φ-OTDR data.