Accurate and robust fault diagnosis of motor bearings plays a vital role in ensuring operational safety and reducing unexpected downtime. A new hybrid deep learning framework is introduced, which integrates symmetric point pattern (SDP) transformation, convolutional neural network (CNN) and bidirectional long short-term memory (BiLSTM) network. This method is used to classify bearing faults in multiple categories by using vibration signals. The 1D raw signals are first converted into 2D SDP images, which enhance the spatial distinguishability of fault patterns. A CNN is then applied to extract deep spatial features, while BiLSTM models the temporal dependencies within feature sequences. Experiments conducted on the benchmark Case Western Reserve University (CWRU) dataset demonstrate that the proposed SDP-CNN-BiLSTM model achieves superior performance with an accuracy of 98.33%, outperforming traditional machine learning models such as RF, SVM, ELM, and CNN only baselines. Visualization through scatter plots and confusion matrices further confirms the effectiveness and interpretability of the method. The results suggest that the proposed framework offers a promising tool for intelligent condition monitoring in industrial applications.