The footstep vibration signals generated during human walking process are highly individualized and can be used as a biological feature for person identification. This study introduces a pedestrian identification method utilizing the Visual Geometry Group Network with 11 layers (VGG11) network, which processes raw time-domain footstep vibration signals through Short Time Fourier Transform (STFT) for frequency analysis. The model is trained on datasets from both flat ground and stairway walking to ensure its robustness in various environments. Comparative analysis reveals that the VGG11 network outperforms traditional classifiers, achieving 96.83% accuracy on flat ground and 94.95% on stairway datasets. Conventional methods like Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) show lower accuracy rates of 90.40% and 84.2%, respectively, on flat ground datasets, and approximately 50% on stairway datasets.