Accurately identifying the lane change intention of vehicles is a key strategy for improving the reliability of driving assistance systems and ensuring road safety. A novel method that combines convolutional neural network （CNN） and gated recurrent unit （GRU）， integrated with a logic judgment mechanism， was proposed to effectively recognize the lane change intentions of vehicles. First， test data from twenty volunteers were recorded using a driving simulator including three categories： left lane change， right lane change， and straight driving. The data was used to construct a sample set of lane change intention. Secondly， a CNN-GRU model was built using vehicle driving characteristics and driver behavior data， with the CNN layer being employed to extract features as input to the GRU layer. Finally， a logic judgment layer was integrated into the intention recognition network to address the temporal dependencies of lane change intentions by setting probability thresholds. To validate the validity of the method in this study， the model was compared and analyzed with a CNN that was integrated with long short-term memory （LSTM） and GRU. The results show that the proposed model achieved recognition accuracies of 98.5% for left lane changes， 96.7% for right lane changes， and 95.2% for straight driving， demonstrating higher accuracy compared with other models.