Arc Sound Recogniting Penetration State Using LPCC Features
Arc Sound Recogniting Penetration State Using LPCC Features
复制标题
DOI：
10.1007/978-3-642-19959-2_28
复制
发表时间：
2011
期刊：
影响因子：
--
通讯作者：
Jifeng Wang;Yantian Zuo;Yi-chang Huang;Bo Yang;S. Pan
中科院分区：
文献类型：
--
作者：
Jifeng Wang;Yantian Zuo;Yi-chang Huang;Bo Yang;S. Pan
焊接是一个复杂的热加工过程，焊接过程中会产生声、光、电、烟尘等，其中蕴含着丰富的焊接质量和焊透信息。随着工业生产的快速发展，对焊接自动化和焊缝精密成形的要求越来越高。焊缝渗透是衡量焊接质量的一项重要指标。对于钨极氩弧焊，利用焊接声信号监测焊池状态具有实际应用的潜力。将类似于语音管道模型的圆弧声道描述为线性预测全极模型。本文在语音识别中采用线性预测倒谱系数技术提取弧声特征，并采用人工神经网络对不同穿透状态进行分类，测试数据准确率达到79%。
Welding is a complex heat working process and it will generate sound, light, electricity, soot and so on, which contain abundant information about welding quality and penetration. As the quick development of industrial production, it is necessary to meet more and more requirements from both welding automation and precise seam forming. The penetration of seam is a common important index of the welding quality. As to Gas Tungsten Argon Welding (GTAW), using welding sound signals to monitor the welding pool state has the potential in the practical application. The arc sound channel which like speech pipeline model was described as a linear prediction all-pole model. This paper adopted linear prediction cepstral coefficients technology in speech recognition to extract the features of arc sound and employs an artificial neural network to classify the different penetration state, with 79% accuracy in test data.