The current research work for picking up and analyzing the pre-shock vibration signals of strong earthquakes is often carried out in a limited frequency band,and the amount of data analyzed has a limited spatial and temporal range. In this paper,we propose a method to calculate and analyze seismic background noise directly based on big data by using clusters for distributed noise power spectrum calculation and storage of calculation results for massive seismic observation data. The continuous waveform data from Sichuan station network from March to April, 2013 was used to analyze the temporal variation of multiple frequency points from 50Hz-200s of recorded data from seismic stations before Lushan earthquake. We found that the MDS and BAX stations within 50km of the epicenter of the Lushan earthquake showed a noise enhancement of about 10～20dB in the horizontal direction for 3～5 days in the 30～150s period in the low frequency band,while no significant changes were found in the vertical direction. This long-period anomaly is difficult to be detected by traditional ground shaking signal pickup,which suggests that the proposed method of strong-motion seismic anomaly analysis is capable of detecting and analyzing long-period anomalous signals.