This paper proposes a method to estimate multiple harmonic families parameters of a signal, which is based on subspace matching pursuit decomposition algorithm, and decomposes the signal into harmonic subspace adaptively. It takes advantages of interpolated FFT algorithm and the prior information of signals parameters, with high estimation precision of harmonic frequency, it has ability to distinguish signals harmonic components of different harmonic families, which have closed-spaced fundamental and harmonic frequencies. The trail indicates that it is a preferable method to estimate multiple harmonic families parameters, as it is insensitive to random phase-shift of harmonic components, and is useful to precisely estimate the fundamental, harmonic frequencies and order of a signal. This method also can be used to depict the features of a ship radiated-noise power spectrum, and to reveal the operation condition of the ship propellers. All of above show the advanced method is a valuable method for multiple harmonic parameters estimation of signals.