In view of such flaws as insufficient precision and limited diagnostic ability under complex fault modes found in bearing fault diagnosis of traditional single mode signal processing, a bearing fault diagnosis method, which is based on convolutional neural network (CNN) and gated recurrent unit (GRU), has thus been proposed. With the time-domain and frequency-domain features of current signals and vibration signals integrated for fault classification, CNN is used for feature extraction, and GRU for capturing the long-term dependencies of time-series data, thus improving the diagnostic ability of the model. In addition, an optimization of the training process and prevent gradient vanishing or exploding can be achieved by adopting batch normalization (BN) method. Experimental results show that the proposed method is characterized with a high classification accuracy in different fault states, especially in normal operation and inner ring fault states, with its accuracy and recall rates close to 1, indicating that the proposed method possesses a strong robustness in the fault diagnosis task of multimodal data fusion. The comparative experimental results with CNN, ResNet, and MS-CNN also demonstrate that the proposed method has a higher accuracy in fault diagnosis under variable speed conditions.