According to the recurrence plot analysis to the gear vibration signal,it is found that there are obvious difference between the gear condition and the driver system characteristics represented by recurrence plots.If the recurrence plots are regarded as a binary image,the parameters calculation by recurrence qualification analysis,such as recurrence rate,determinism,laminarity and recurrence time etc can be used to describe the feature of the image.The characteristics of recurrence plots that correspond to dynamical system and image respectively are analyzed.In order to depict the local feature of recurrence plot,the sliding recurrence analysis is proposed.The proposed approach combined with Gaussian mixture model and Bayesian maximum likelihood classifier are used to classify the gear vibration signals which are acquired from gear fault experiment facility.The classification results show that the higher discrimination rate can be achieved by the proposed method.