Abstract: To address the difficulty of identifying cutting medium properties in harsh mining environments using single sensor signals for the roadheader, a multi-information fusion recognition method based on sound and vibration signals is proposed. First, to tackle the problem of noise interference in underground environments, the VMD algorithm is used to decompose the sound and vibration signals. A modal selection criterion based on Shannon entropy is proposed to reconstruct and denoise the signals. Next, based on the Markov transition field, the features of the sound and vibration signals are converted into two-dimensional images for feature fusion. Finally, a convolutional neural network (CNN) combined with a convolutional block attention module (CBAM) is introduced to perform spatial feature learning on the two-dimensional image information, enabling automatic extraction of multi-scale features and precise enhancement of key features, thus achieving accurate identification of the cutting medium properties. Experimental results show that the proposed method achieves an identification accuracy of 99.4%, outperforming other traditional methods. This method provides a theoretical foundation for the construction of intelligent mines and unmanned mining operations in the excavation face.