Research on heart rate anomaly detection model for coal miners based on multi-scale feature fusion

To reduce accidents caused by abnormal heart rates among coal miners, this study investigates a method for identifying and determining abnormal heart rate patterns using deep learning techniques. By comparing and analyzing the performance of three models—convolutional neural network (CNN), long short-term memory network (LSTM), and multi-channel convolutional neural network (MCNN)—in classifying abnormal heart rate patterns, the advantages and limitations of each model in handling complex time-series data are explored. Experimental results indicate that the CNN model achieves approximately 97% accuracy on the training set and around 88% on the validation set, demonstrating its capability to effectively capture feature patterns in ECG signals. However, its generalization ability is slightly inferior to that of the LSTM and MCNN models. The LSTM model achieves over 98% accuracy on the training set, with the validation accuracy stabilizing around 89%, indicating stronger generalization capability. The MCNN model attains over 98.5% accuracy on the training set and 90% on the validation set, demonstrating excellent classification performance and strong robustness, particularly in handling complex data distributions. In summary, the proposed deep learning models can accurately detect changes in heart rate and timely identify abnormal states, providing an effective technical approach for health management and accident prevention in coal mining operations.