工业现场强噪声环境下目标声音片段的提取

Method for impact sound extraction in industrial noisy environment

音频法作为一种低成本无损检测手段应用于工业在线检测时,工业现场的强噪声给目标声音信号的特征提取带来极大困难。给出一种强噪声环境下的目标声音提取方法,该方法设计了自屏蔽式声音采集装置;在分析现场实测音频信号成分基础上,提出通过机电结合方法的声音激励初次预测技术;运用短时傅里叶变换的目标声音时频域二次精确定位算法,实现了工业现场强噪声环境下目标声音片段的精确提取。实验结果表明:在2 m范围内自屏蔽声音采集装置可将噪声源声强降低20 dB上;声音片段提取的时间精度优于10 ms。

While sonic detection is used in industrial on-line detection as a non-destructive, low-cost method, the high-level noise in working place makes the target-sound feature extraction very difficult. A novel method for target-sound extraction in noisy environment is presented in the paper, which designs a sonic acquisition device with self-shielding, puts forth the first prediction technique of sonic stimulation based on the sound signal component analysis in actual measurement and the second location algorithm that transforms signal from time domain to time and frequency domain via fast Fourier transform （FFT） to realize accurate extraction of target-sound signal in industrial noisy environment. Experiments show that the interference of noise source can be lowered by 20 dB or more and the accuracy of extraction time is better than 10 ms.