Based on the MATB II experimental platform, flight test simulation tasks under three different mental loads were designed, and electrocardiographic (ECG) data were collected simultaneously. ECG signals are susceptible to noise from different sources, which increases the possibility of misreading and may affect the heart rate variability (HRV) characteristic index calculation process. Aiming at the noise of irregular mutation in the waveform of the ECG signal, this paper proposes a fast Fourier transform (FFT) to calculate the correlation coefficient of the normalized value of the unilateral spectrum between adjacent windows of the ECG signal, and set a threshold to eliminate The method of irregular mutation data segment. The results show that the algorithm has a significant improvement in removing noise from ECG signals. By comparing the abnormal judgment results of human eye observation and calculation of frequency-domain correlation coefficient, the judgment threshold R = 0.95 with the most similar rejection effect is obtained. This method can be used to denoise ECG data under different mental loads.