抄録
This paper describes a high-resolution frequency identification method based on complex spectral analysis. In power spectral analysis, the resolution of frequency in periodic wave is limited by 1/T, where T is the observasion interval. When the two nearby frequencies produce a single peak in power spectrum, they can not be identified separately. But the phase component in complex spectrum gives us a useful method to increase the resolution of frequency identification. Complex spectrum of a periodic wave has the following characteristics as: 1) The direction of component-vector is identical and determined by original wave phase φ. 2) The inverse of amplitude of component-vector gives a linear function of component-frequency. And 3) the cross point of the linear function and the frequency axis gives the frequency of periodic wave. This paper proposes a method to determine whether the peak consists of one periodic signal or two-frequency signals, and to identify the accurate frequencies contained in the original wave signal.
Experimental result shows that two nearby frequencies in original signal can be accurately identified from a short-time complex spectrum. And the influence of the sidelobe of the other spectral peak on the identification accuracy is also discussed.
