Traditional frequency self-adjusting algorithm calculates the frequency spectrum directly

which will affect the noise extraction of the signal.Logarithm power spectrum is used to guarantee the positive-value characteristic of the power spectrum

and make the smoothing algorithm with forgetting factor easier to implement.To solve the contradiction between convergence speed and convergence precision of the fixed forgetting factor

a self-adjusting method with varied forgetting factor is proposed. The convergence condition is also deduced. The iteration algorithm whose forgetting factor is the exponential mean value of the response spectrum is designed.Simulation results show that the proposed spectrum equalization algorithm reduces the calculation time by 50%

the steadystate error by 90%

comparing to un-optimized algorithm.