The precision of vibration frequency estimation by adaptive notch filter (ANF) is not satisfactory all the time. The analysis pointed out that the fundamental reason is the ANF error converges to a local optimum value. A feedback corrected ANF is proposed referring to the idea of feedback control. According to the correlation between the input signal and the output signal, a factor evaluating the frequency tracking accuracy is designed to monitor the ANF work status. If the factor is less than a given threshold value, it’s considered the ANF lost vibration frequency. Then, it is demanded to enlarge the bandwidth of the ANF, so that the frequency can be re-tracked accurately. To insure the re-converging speed, a strategy of feedback adjusting the ANF parameters appropriately rather than re-initialization is applied. The ANF based on Steiglitz-McBride method(SMM-ANF) is tacked as an example to illustrate the feedback correction strategy. The signal processing flow and particular algorithm are also exhibited. The performance of Lattice ANF, SMM-ANF and the feedback corrected SMM-ANF are compared by simulations. An application instance of the proposed method in coriolis mass flow meter is given out briefly. Results show that the convergence rate of the feedback corrected SMM-ANF is little lower than original SMM-ANF, higher than the lattice ANF. The continual frequency tracking performance of the feedback corrected SMM-ANF is obviously better than the others.