The MEMS seismograph developed based on Micro-Electro-Mechanical System (MEMS) technology has the advantages of easy integration, low maintenance cost and low power consumption, and is widely used in the field of seismic monitoring. However, the integrated software and hardware resources of MEMS seismograph are limited, and it is greatly interfered by the instrument's own noise and other factors, resulting in low-quality seismic signal measurements, which requires higher embedding algorithm. To solve this problem, this paper proposes an improved Short Term Average/Long Term Average (STA/LTA) algorithm that is more suitable for MEMS seismographs. Firstly, the characteristic function of Anti-interference (AR) is constructed to suppress the interference of baseline drift and low-frequency noise, and improve the anti-interference ability of the STA/LTA algorithm to pick up seismic events. Secondly, the method of "delay time window" is proposed to improve the computational efficiency and picking accuracy of the STA/LTA algorithm, and reduce the occupation of MEMS integration resources by the STA/LTA algorithm. Finally, the influence of different time window sizes on the picking efficiency of STA/LTA algorithm is further explored by combining the time window position. The simulation results of actual seismic data show that the improved STA/LTA algorithm proposed in this paper has higher computational efficiency, stronger real-time and anti-interference ability, and is more suitable for integrating MEMS seismometers with limited resources.