Spectrum feature extraction plays a crucial role in identifying earthquake events and calculating seismic parameters. However, the identification standards of effective modal components in variational mode decomposition (VMD) are incomplete, leading to imprecise spectrum feature extraction. To address this issue, we propose a novel seismic spectrum feature extraction method that combines Allan variance, VMD, and power spectral density (PSD). First, VMD is applied to filter noise components from triaxial accelerometer observations and add effective signals. Second, PSD is utilized to extract three groups of seismic frequencies (tri-axial accelerometers). Finally, the Allan method is introduced to identify the group of accelerometer observations with the highest reliability as the vibration frequency caused by the seismic excitation. We validate the effectiveness of our method by analyzing a Mw 2.6 microseismic event that occurred in Huairou, Beijing in 2022. Our analysis demonstrates that triaxial accelerometers can effectively detect such events with a magnitude of 2.6. Additionally, our proposed method accurately extracts seismic spectrum features. Specifically, the seismic excitation vibration frequencies at four seismic monitoring stations were found to be 26.95 Hz, 12.89 Hz, 12.89 Hz, and 12.5 Hz. These findings highlight the potential of our approach for identifying and characterizing earthquake events, which has important implications for earthquake monitoring and hazard assessment.