It is important to detect defects in wood, when it reduce the performance. The data and signal processing technology providing researchers with more damage identification problem solution ideas and methods. This article explore the wavelet analysis and artificial neural network for the wood defects based on non-destructive testing, and build an artificial neural network model for wood non-destructive testing technology. After wavelet packet decomposition to extract the different frequency bands of energy levels characteristic of the signal, as the neural network input samples, the network training and learning. Training of the BP network model can be achieved on the different locations automatic recognition of defects, defects of the middle of more than 90% recognition rate on the left and right side of the recognition rate of over 80%.