Empty Diagnosis for Wood Based on Wavelet Packet and RBF Neural Network

The health Mongolian Oak specimens and the Mongolian Oak specimens with empty were studied by the method of wavelet packet and RBF neural network loosely bound. The wavelet packet transform was used to analyze the stress wave detection signal to carry on the 5layer wavelet packet analysis, and the 8 dimensional feature vector was constructed. Then the feature vector was used to train the RBF neural network and establish the diagnosis model. The experimental result showed that the identification accuracy rate of the model was up to 9080%, which could effectively evaluate the nature of wood, and a theoretical reference for the design of stress wave nondestructive testing instrument was provided.