This study proposes a method to indicate the optimal moment to interrupt grinding wheel dressing using the counts statistics extracted from acoustic emission signals. Experiments were conducted on a surface grinding machine equipped with an aluminum oxide grinding wheel and a single-point diamond dresser. A Multilayer Perceptron neural network, trained via the Scaled Conjugate Gradient algorithm, classified the grinding wheel condition into three states: undressed, intermediate, and dressed. In three external tests not included in the training set, the model achieved an accuracy of 88.13%.