Dual-Feature-Based Bubble Sound Detection Method and Its Application in Passive Acoustical Detection of Underwater Gas Leakage

Detecting acoustical signals arising from underwater gas leaks is crucial for monitoring greenhouse gas emissions from submarine vents using passive acoustical monitors. When gas bubbles intermittently generate, direct detection of the sound signals produced by these bubbles is an effective method for identifying underwater gas leaks. However, traditional energy detectors lack the capability to specifically detect bubble sound signals, making them susceptible to interference from marine environmental noise. Through an analysis of instantaneous bandwidth variation, we have identified two distinct feature components of bubble sound signals: short-term harmonic and wideband pulse. To address this, this article introduces a dual-feature-based bubble sound detection method. The method includes a bubble sound detector employing a sparse morphological component analysis (MCA) algorithm designed to extract these two feature components in both the time domain and time-frequency domain. The proposed feature-based detector demonstrates reliability and robustness against impulsive noise within ocean ambient noise. Furthermore, the proposed feature-based detector is applicable to the binary classification task of gas leak detection. Experimental results confirm the reliability and robustness of the proposed method in detecting underwater gas leaks.