This paper proposes a biased frequency adaptive comb filter which is composed of a first-order low pass filter and multi normalized frequency estimators in parallel. The algorithm is used to estimate fast and precisely the bias component and the frequency and the amplitude of each harmonic/inter-harmonic component in the sum of direct current component and multi harmonic components. The bandwidth parameter and the frequency adaptive gain give different effects on the convergence speed and the precision of frequency estimation and amplitude estimation respectively. A discrete algorithm is obtained by us of classic fourth-order Runge-Kutta method, the program flow for realization of which is presented in this paper. The experiment results from an experimental system composed of two DSPs verify the validity of the proposed method.