Recognition of Wood-Floor Damages Based on Wavelet Transform and Neural Network Ensemble

Analyzed the forced vibration dynamic characteristics of three different damage-types wood-floors, according to the characteristics of forced vibration signals, wavelet packet decompose was proposed to extract the information related to the condition of the wood-floor materials from the data and served as characteristic parameters to be putted into neural network ensemble. The different damage-types of wood-floor can be recognized by artificial neural network ensemble if the reasonable artificial neural network ensemble model was chosen. The results show that the method of extracting the feature and the neural network ensemble model are effective for identifying the wood-floor damages. And, the recognition of the neural network ensemble is more accurate than that of single network classifier for wood-floor damage.