The invention discloses a decision feedback equalization algorithm with low complexity. The algorithm comprises the steps of performing a frequency domain transformation on receiving signals at a receiving end, performing a linear equalization and transforming the signals to a time domain, and inputting the signals into a noise predictor to perform noise prediction after the signals are decided. Thus, the noise interference of un-decided symbols is reduced, and the performance of a communication system is improved. When the coefficient of the noise predictor is computed, a threshold value anda feedback order are firstly set; part of coefficient values of a feedback equalizer is particularly computed by the prior noise prediction decision feedback algorithm; then, a prediction model is built according to channel characteristics; the rest of feedback coefficients are predicted by utilizing the acquired coefficients of the feedback equalizer; and thus, the computation complexity is reduced.