基于小波包变换的超声回波信号特征提取

Feature extraction of ultrasonic echo signals based on wavelet packet transform

超声检测技术中,缺陷的定性分类这一技术难题至今尚未得以彻底解决。文章介绍了一种基于小波包变换的多尺度空间能量特征提取方法,并对实测的超声缺陷回波信号进行了特征提取;进而采用基于距离的类别可分性判据对提取结果进行了评价。结果表明:该方法提取的缺陷回波信号特征值,其可分性测度均值达91.7%,从而证明该方法对超声检测缺陷回波信号的特征提取是相当有效的。

The problem of flaw classification in ultrasonic testing has not been solved so far. In this paper, a method of feature extraction of ultrasonic flaw echo signals is introduced, which is based on the theory of multilevel wavelet packet transform. The characteristic values of measured echo signals are extracted by this method and evaluated by the separability criterion. The average value of separability measure is as high as 91.7%, which indicates that this method is quite effective in the feature extraction of ultrasonic flaw echoes.