The invention discloses a digital background correction method based on a least mean square algorithm, which is suitable for a split SAR ADC. The method comprises the following steps: firstly, settinga split SAR ADC, and performing digital background correction based on a minimum mean square algorithm on the set split SAR ADC, wherein the split type SAR ADC comprises two ADC modules, a main DAC capacitor array in each ADC module adopts a non-binary capacitor array; meanwhile, the weight of the highest-order capacitor in the main DAC redundant capacitor array is set to be minimum, so that eachcapacitor including the capacitor with the maximum weight in the main DAC capacitor array can be effectively corrected through random switching of the correction DAC capacitor array, and the linearity and the dynamic range of the ADC are improved; the introduction of the correction DAC random switching mode can effectively solve the problem that the correction is invalid due to the fact that thetwo ADC capacitors are not consistent in direction. In addition, redundancy is introduced into the main DAC capacitor array, so that dynamic errors introduced into the quantization process of the system can be weakened, the correctness of each switching is ensured, and the iteration speed is increased.