基于声音特征的隧道衬砌空洞识别方法研究

Research on tunnel lining cavity recognition method based on acoustic characteristics

目前隧道衬砌空洞检测以人工敲击判断为主,检测过程中由于受到检测人员水平、注意力等主观因素影响,检测结果存在较大不确定性,因此有必要研制一种智能化的检测装置实现空洞自动识别。文章开展了衬砌空洞敲击回声智能识别算法研究,通过提取隧道衬砌冲击回波的梅尔倒谱系数(Mel Frequency Cepstral Coefficient,MFCC)作为特征,针对敲击回声脉冲信号长度不一的特点,提出了变帧长MFCC优化算法,并面向小样本条件,建立了支持向量机(Support Vector Machine,SVM)的识别模型。试验结果表明,该模型对衬砌空洞识别准确率可达89.9%。

At present,the detection of tunnel lining cavities is mainly based on manual knocking judgment.The detection process is affected by the professional level,attention and other factors of the testing personnel,which brings more uncertainty to the detection results.Therefore it is necessary to develop an intelligent detection device.In this paper,the intelligent recognition algorithm of the echo of knocking lining cavity is studied.The Mel cepstrum coefficient(MFCC)of the acoustic signal is extracted as the feature.A MFCC optimization algorithm is proposed in view of the different length of echo pulse signals,and the recognition model based on the support vector machine(SVM)is built under the condition of small dataset.The test results show that the accuracy rate of the lining cavity identification of this model can reach 89.9%.