In order to solve the diamond roller dressing wear situation is difficult to accurately online judgment problem. A set of built-in acoustic emission on-line monitoring diamond roller wear device was designed and developed. Based on Shannon information entropy theory and wavelet packet model, a method of acoustic emission signal processing for diamond wheel dressing was proposed. The method enabled the information entropy of wavelet packet coefficients to be calculated for each layer, and the optimal number of wavelet packet decomposition layers could be determined according to the variation rule of information entropy. Then, the wavelet packet decomposition feature parameter was downsized using principal component analysis, and the feature parameter that best characterizes the wear of diamond rollers was extracted. Finally, PSO-SVM (particle swarm optimization algorithm support vector machine) based on this feature parameter was established. The wear state of diamond rollers was monitored experimentally. The results show that the PSO-SVM model has the highest classification accuracy, with an average correct rate of more than 95. 24% , and the effectiveness of PSO-SVM is verified by a large number of experiments.