Abstract: A new wavelet analysis method for acoustic emission tool wear distinguish is presented. The local characterize of the frequency band which contains the signal main energy is depicted by the multiplayer wavelet decomposition. Since then, the physical relationship between the wavelet decomposition coefficients’ statistical value and tool state can be built up. Its validity is proved by the turning experiments. It has much more theory visualizability and maneuverability than the usual prediction method which based on artificial nerve network(ANN).