针对金属板下多层非金属粘接结构试件中的超声检测回波信号，提出了一种基于小波包变换的提取特征参数的方法，以评估试件的粘接特性。在利用超声信号进行层状粘接结构的无损评价时，各层的超声检测回波会严重混迭，难以分辨甚至会丢失某一层的缺陷信息。利用信号处理技术，以小波包变换后的能量来进行深层界面的特征提取。结果显示，提取出的特征可有效地进行各层粘接情况的分析和识别。试验检测结果表明特征提取方法可用在金属基多层材料结构的无损检测中。

This paper present a wavelet-packet-based approach for extracting feature from ultrasonic non-destructive evaluation data of a test multi-layer structure specimen. These features are then used to obtain an estimate of the specimens bonded properties. When the multi-layer structure undergo ultrasonic non-destructive evaluation, echo from interior layers often overlap severely, causing information from individual layers to be obscured or lost completely. A complex-valued mother wavelet based on the impulse response of the test instrumentation system, is used to perform the wavelet transform and separate the overlap echoes. The extracted features displayed are effective to analyze and identify the bonded properties. It is believed that the features extracted via the method outlined in this paper provides considerable promise in the area of the test of multi-layered bonded structure.