Browsing by Author "A Ismail"
  • Publication
    Characterization of DWT as denoising method for φ-OTDR signal
    ( 2021-12)
    M S Yusri
    ;
    B Faisal
    ;
    A Ismail
    ;
    N L Saleh
    ;
    M F Ismail
    ;
    N D Nordin
    ;
    A H Sulaiman
    ;
    F Abdullah
    ;
    M Z Jamaludin
    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.
      1  1
  • Publication
    Improving event classification using Gammatone filter for distributed acoustic sensing
    ( 2021-12)
    B Faisal
    ;
    M S Yusri
    ;
    A Ismail
    ;
    N L Saleh
    ;
    M F Ismail
    ;
    N D Nordin
    ;
    A H Sulaiman
    ;
    F Abdullah
    ;
    M Z Jamaludin
    The phase optical time domain reflectometry (Φ-OTDR) system offers several advantages suitable for distributed acoustic sensing application. It has long sensing range, great anti-electromagnetic interference, and high sensitivity towards environmental vibrations. However, as a sensor system, the Φ-OTDR is limited to only collecting environmental vibrations without providing more useful information such as the location and types of events happening around the sensing region. Therefore, it requires an extensive data processing system to distinguish between different events happening within the sensing regions. In this paper, Simple Differential and Normalized Differential method were used to extract perturbation event prior to classification process comprising data organization, features extraction, and classification outcome were implemented. Gammatone Frequency Cepstral Cepstrum were used to handcraft features for classification and were obtained using Gammatone Filter processing. Classification scheme based on Support Vector Machine (SVM) is use as classifier where accuracy score 100%.