This paper presents a hybrid CNN-LSTM neural network model for predicting the seismic response of high-speed railway tracks. The model uses quasi-distributed fiber Bragg grating (FBG) sensors to monitor track responses during seismic events. Seven grating points are placed on each fiber to capture responses of the track plate, rail, base plate, and beam. Data from peripheral gratings are used to predict the central point's response. A continuous feature map is created using a time-sliding window from the rail's acquisition location, which undergoes initial feature extraction with CNN. These features are then sequenced for the LSTM network for prediction. The model achieves an RMSE of 0.3753, MAE of 0.2968, and R² of 0.9371, demonstrating its effectiveness in earthquake response analysis for rail infrastructure. High-speed railways have significantly expanded in China, and the seismic performance of railway bridges and tracks is crucial for safety. Researchers have studied the seismic performance of simply supported girder bridges and CRTSII track slabs, finding that stresses in rails, track plates, and base plates are present near bridge abutments or anchorages. Various studies have explored the seismic behavior of reinforced concrete bridge piers and the stochastic seismic response of high-speed rail systems. Shaking table tests have been used to simulate earthquake vibrations and study structural responses. FBG sensors have been widely used to assess structural conditions of existing infrastructure. Recent studies have explored the potential of deep learning in seismic response prediction, including the use of CNN and LSTM networks for time series prediction. This research proposes a CNN-LSTM hybrid model to improve the accuracy of seismic response prediction for high-speed railway tracks. The model combines CNN and LSTM features, uses quasi-distributed FBG sensors to collect stresses on simply supported girder bridges under seismic impacts, and creates a continuous feature map of observed grating locations, seismic orientations, and peak accelerations as input. The model is capable of predicting strain across the span, showing potential for improved accuracy in seismic response predictions. However, the computational effectiveness of the model in real-life scenarios, especially for high-speed railway bridges, requires further exploration.This paper presents a hybrid CNN-LSTM neural network model for predicting the seismic response of high-speed railway tracks. The model uses quasi-distributed fiber Bragg grating (FBG) sensors to monitor track responses during seismic events. Seven grating points are placed on each fiber to capture responses of the track plate, rail, base plate, and beam. Data from peripheral gratings are used to predict the central point's response. A continuous feature map is created using a time-sliding window from the rail's acquisition location, which undergoes initial feature extraction with CNN. These features are then sequenced for the LSTM network for prediction. The model achieves an RMSE of 0.3753, MAE of 0.2968, and R² of 0.9371, demonstrating its effectiveness in earthquake response analysis for rail infrastructure. High-speed railways have significantly expanded in China, and the seismic performance of railway bridges and tracks is crucial for safety. Researchers have studied the seismic performance of simply supported girder bridges and CRTSII track slabs, finding that stresses in rails, track plates, and base plates are present near bridge abutments or anchorages. Various studies have explored the seismic behavior of reinforced concrete bridge piers and the stochastic seismic response of high-speed rail systems. Shaking table tests have been used to simulate earthquake vibrations and study structural responses. FBG sensors have been widely used to assess structural conditions of existing infrastructure. Recent studies have explored the potential of deep learning in seismic response prediction, including the use of CNN and LSTM networks for time series prediction. This research proposes a CNN-LSTM hybrid model to improve the accuracy of seismic response prediction for high-speed railway tracks. The model combines CNN and LSTM features, uses quasi-distributed FBG sensors to collect stresses on simply supported girder bridges under seismic impacts, and creates a continuous feature map of observed grating locations, seismic orientations, and peak accelerations as input. The model is capable of predicting strain across the span, showing potential for improved accuracy in seismic response predictions. However, the computational effectiveness of the model in real-life scenarios, especially for high-speed railway bridges, requires further exploration.