Abstract: In the advanced detection of coal mine excavation face, the detection technology for identifying abnormal bodies has limitations, resulting in insufficient accuracy. Traditional detection methods, such as certain physical detection techniques, are relatively limited in their detection range and accuracy due to the limitations of tunnel space. When dealing with complex geological conditions, it is difficult to accurately identify the location and shape of anomalous bodies. To this end, research is being conducted on recognition methods based on improved convolutional neural networks (CNN). By preprocessing the long-distance advanced geophysical data of sound waves, including steps such as denoising, enhancement, and normalization, the data quality is improved. Using an improved CNN based model to extract anomalous body features from detected images, this model effectively improves the accuracy and robustness of feature extraction by optimizing convolutional layers, introducing attention mechanisms, and adjusting hyperparameters. Finally, based on the extracted feature vectors, an SVM classifier is used to recognize and classify anomalous bodies. Through comparative experiments, it has been proven that this method significantly improves the accuracy and efficiency of anomaly recognition compared to existing methods.