基于AR-SK的北美鹅掌楸原木声信号特征参数提取及质量分等  

A feature extraction and quality grading method of acoustic signals generated from yellow poplar(liriodendron tulipifera)logs based on AR-SK

作　　者：徐锋[1] 瞿玉莹 XU Feng;QU Yuying(College of Information Science and Technology,Nanjing Forestry University,Nanjing 210037,China)

机构地区：[1]南京林业大学信息科学技术学院,南京210037

出　　处：《振动与冲击》2020年第24期99-106,149,共9页Journal of Vibration and Shock

基　　金：国家自然科学基金项目(31170668);江苏高校优势学科建设工程资助项目;南京林业大学优秀博士学位论文创新基金项目(163070682);2017年度大学生实践创新训练计划项目(201710298028Z)。

摘　　要：阔叶材原木的质量评估可以为行业提供最优的阔叶材资源利用价值,然而因缺陷声信号的非平稳性和缺陷类型特征的重叠性,有效的质量评估声参数非常有限。基于此,提出一种基于自回归模型(AR)和谱峭度(SK)相结合的特征声参数提取与分等方法。利用AR线性滤波器滤除声信号中周期平稳成分,并对包含缺陷信息的残差信号进行短时傅里叶变换,计算其谱峭度值并定位最大谱峭度所在的频带,以其频带的中心频率和带宽设计滤波器对残差信号进行滤波以获取原木主要缺陷信号分量,计算该信号分量的峭度值,将其作为表征声信号的特征参数对北美鹅掌楸原木进行质量分等。样本原木的实际锯切结果显示,基于AR-SK的预测分等中,高质量原木组中的高等级板材率为77.2%,而低质量组中的高等级板材率为21.8%。与传统的声速分等相比,高质量组中的高等级板材率提高了33%以上,而低质量组中的高等级板材率降低了约26%。研究结果表明,所提方法能有效分离缺陷信号成分并对该原木质量进行较精确分等。Aiming at the insufficient acoustic parameters in quality assessment of hardwood logs due to the non-stationary features of acoustic signals and the overlapping of defect features,a method for feature extraction and quality grading was proposed based on the autoregressive model(AR)and spectral kurtosis(SK).An AR-based linear filter was applied to filter periodic deterministic components from original signals according to the Akaike information criterion,and the residual signal containing the defect information was decomposed by the short-time Fourier transform for acquiring the SK values of sub-band components.Then,a filter was designed to filter the residual signal further for obtaining the main defect components according to the center frequency and bandwidth of the sub-band where the maximum SK value located in.Finally,the kurtosis of the filtered defect signal was used as the characteristic parameter for quality grading of the yellow poplar logs.The sawing results of the sample logs show that for the predicted log quality grade results based on AR-SK,the high-grade boards occupancy is 77.2% in the high quality log group,but that is 21.8% in the low quality group.Compared with the log quality’s predicted grade results based on the traditional velocity,the high-grade boards occupancy is increased by more than 33% in the high quality log group,while that is decreased by about 26% in the low quality log group.The results show that the proposed method can effectively separate the defect signal components from the original signal and accurately classify these hardwood logs in quality.

关 键 词：阔叶材原木 质量分等 峭度 谱峭度(SK) 自回归模型 

分 类 号：TB529[理学—物理]