The steam generator (SG) is a heat exchange device between a sodium coolant and water in sodium-cooled fast reactors. Its transfer tubes are prone to leakage to result in sodium-water reaction, which will affect the operational safety of the entire nuclear reactor. To this end, a signal processing method is studied to detect the SG leakage using an electromagnetic vortex flowmeter (EVFM). Water flow is firstly used to replace liquid sodium for conducting the hydraulic experiments, and the output signal of the sensor of the EVFM (abbreviated as EVFM sensor output signal) is collected under different gas injection volumes. The characteristics of the signal are analyzed in the time domain, and the collected signals are modeled. A signal processing method of bubble detection in liquid sodium is proposed based on the peak-to-peak standard deviation, which is realized in real time on the hardware system based on a digital signal processor. The gas injection experiments in liquid sodium are conducted to verify the effectiveness of the signal processing method and system developed in this paper. The minimum amount leakage of water can be detected as 0.1 g/s.