以少量資料進行自來水管漏水聲偵測​

Abstract

音訊分類技術在各種應用領域中皆具有重要性，尤其在異常事件偵測方面更扮演關鍵角色。本研究與旭儀企業股份有限公司合作，針對自來水管漏水聲之偵測問題進行探討。我們提出一種結合原型網路（prototypical network）與 VGG11 圖像分類模型的架構，並進一步引入本研究所設計的改良方案，以提升分類效能。為驗證所提方法之準確性與穩定性，我們採用公開音訊分類資料集 ESC-50 進行實驗。實驗結果顯示，本方法在處理漏水聲音訊分類任務上具備良好之辨識能力，顯示其應用潛力。

Audio classification plays a critical role in various application domains, particularly in the detection of abnormal events. In this study, we collaborated with ASAHI SUNRISE CO., LTD. to investigate the problem of detecting water leakage sounds in water pipelines. We propose a method that integrates the prototypical network framework with the VGG11 image classification model, along with enhancements specifically designed for this task. To evaluate the accuracy and robustness of the proposed approach, we conducted experiments using the publicly available ESC-50 audio classification dataset. The experimental results demonstrate that our method achieves promising performance in detecting water leakage sounds, indicating its potential for practical applications.