Abstract: The Chang'e-4 low-frequency radio spectrometer is placed on the back of the moon, and its natural observation conditions are unique. However, the Chang'e-4 platform has strong interference of about 10^-15W/(m²·Hz) level, and there are obvious differences in the interference in each channel of time domain data, which greatly weakens the astronomical observation sensitivity of the low-frequency radio spectrometer. To this end, starting from the correlation of the two sets of signals, this paper proposes to divide the time-domain observation data of the low-frequency radio spectrometer A, B, and C antennas into strongly correlated CLEAN model signals and partial correlated residual signal based on the CLEAN algorithm, with the help of cross-correlation power spectrum, Fourier series expansion and other tools. The CLEAN model signal is mainly composed of platform interference signals and possible low-frequency strong radio bursts; the residual signal is composed of receiver noise, undeducted platform interference signals and conventional low-frequency radio signals. Applying this method to actual data, the results show that the unintegrated sensitivity of the Chang'e-4 low-frequency radio spectrometer can be increased by about 8 orders of magnitude, reaching a level of 10^-23W/(m²·Hz). On this basis, based on the separate processing of the deterministic components and broadband random components in the platform interference signal, and then relying on the different performance of the low frequency radio burst signal and the platform interference signal in the power spectrum, and the conventional low frequency radio astronomy signal modulated by the moon rotation and other information, the focus of future scientific analysis work is to further process the CLEAN model signal and residual signal, with a view to discovering low-frequency strong radio astronomy burst signals, and even roughly imaging the entire sky.