Improving event classification using Gammatone Filter for distributed acoustic sensing

Abstract

The phase optical time domain reflectometry (Φ-OTDR) system offers several advantages suitable for distributed acoustic sensing application. It has long sensing range, great antielectromagnetic 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%.