The wear of grinding wheel needs repeated dressing in the process of grinding. The monitoring of grinding wheel wear state can effectively distinguish the working state of grinding wheel and reduce the number of grinding wheel dressing. An monitoring model of grinding wheel wear based on acoustic emission signal is established, and a method of grinding wheel wear state feature extraction based on statistical analysis of wavelet decomposition coefficient mean square value is proposed. At the same time, BP neural network is used to identify the wear state of grinding wheel, which input is three kinds of extraction features and the output is three different grinding wheel wear states. The monitoring system is evaluated by grinding tests. The results show that the proposed acoustic emission signal feature extraction method based on the statistical analysis of average wavelet decomposition coefficient and the monitoring system of grinding wheel wear have good results.