Selecting feature in slow-speed and heavy-load equipments has always been a difficult problem. A new feature selection method based on Laplacian Score is used to Acoustic Emission signal. The more capable of describing the sample clustering property, the more important the selected feature is. The method is a ‘filter’ and unsupervised feature selection method which is just dependent on the space distribution of the sample instead of classifier. Therefore, the method enjoys a simple algorithm and low complexity. The effectiveness of the method is verified by the AE datasets from the bearings of a blast furnace’s belt conveyor.