小波变换时频特性的信号识别

Signal Recognition of Time Frequency Property of Wavelet Transform

研究了小波变换时频特性的信号识别及其在复合材料损伤检测中的应用．根据小波变换的框架重构理论及时频相空间理论，提取信号的时频域特征，通过比较原信号的时频空间和小波变换相空间的相同部分，得到能反映同样时频特征的小波级数展开项和的个数，并用误差函数的最小化提取能反映时频性质的小波系数．以此作为小波神经网络的学习参数，经过学习后，使之能对信号进行识别．应用此方法对复合材料试验过程中的复杂曲线进行了实验识别，效果很好．从小波时频特性提取的信号特征，在时间和频率方面都能体现原信号所包括的本质信息，供助Ｂ样条小波神经网络的识别结果，达到了预期目的．

HJ*4/9〗The signal recognition of time frequency property of wavelet transform and its application in the damage detection of composite materials were studied. According to the reconstruction of frame of wavelet and the theory of phase space in wavelet transform, features of the signal in the field of time frequency were extracted. We obtained the terms of expansion of the wavelet series which have the same property as the original signal by comparing the time frequency space of the signal with phase space of the wavelet transform. We extracted the feature coefficients of the wavelet series by computing the minimum mean squared error based on the idea of the gradient algorithm. The features then can be put to the adaptive wavelet network for training and classifying. It shows good performance of applying the concepts to the damage detection of composite materials. It is concluded that the recognition of signal based on time frequency property of wavelet and B spline wavelet neural network gives good results. The experiments demonstrate the feasibility of the proposed method.