【摘要】 针对石油化工等复杂高危场所，危化品泄漏产生的非平稳泄漏声音信号，难以正确判断声音端点的问题，基于小波和经验模态分解（EMD,Empirical Mode Decomposition），提出一种改进的能零比气体泄漏声音端点检测算法。首先，通过麦克风阵列采集气体泄漏信号，将预处理后的泄漏信号通过小波阈值去噪，以提高检测信号的信噪比；其次，利用EMD算法对降噪的信号进行分解，从分解得到的本征模态分量（IMF,Intrinsic Mode Function）中选择IMF，构造新的信号；然后，通过分帧和加窗的方法对重构信号进行再处理，计算出各帧信号的能零比值，采用提出的自适应门限计算方法对信号进行端点检测；最后，搭建简易的气体泄漏模拟实验平台，对改进的气体泄漏声音端点检测算法，进行了实验测试。实验结果表明：改进的能零比气体泄漏声音端点检测算法，在低信噪比的条件下，仍然具有良好的检测精度和检测效率；与传统方法和基于EMD的能零比算法相比更接近实验采集声音信号真正的声音端点。更多还原

【Abstract】 In order to solve the problem of being difficult to determine the sound endpoint of non-stationary leakage sound signals produced by dangerous chemicals in complex and high-risk places such as petrochemical industry, an improved energy-zero ratio gas leakage sound endpoint detection algorithm was proposed based on wavelet and Empirical Mode Decomposition(EMD). Firstly, gas leak signals are collected by microphone array, and the pre-processed leak signals are denoised by wavelet threshold to improve the signal-to-noise ratio of the detected signals. Secondly, EMD algorithm is used to decompose the denoising signal, and IMF is selected from the Intrinsic Mode Function(IMF) to construct a new signal. Then, the reconstructed signals were reprocessed by means of frame splitting and window adding, and the energy zero ratio of each frame was calculated. The endpoints of the signals were detected by the proposed adaptive threshold calculation method. Finally, a simple gas leakage simulation experimental platform was built, and the improved gas leakage sound endpoint detection algorithm was tested. The experimental results show that the improved algorithm still has good detection accuracy and efficiency under the condition of low signal-to-noise ratio. Compared with the traditional method and the EMD-based energy-zero ratio algorithm, it has been closer to the real sound end point of the experimental acoustic signal acquisition.更多还原