INTERPOLATION TECHNIQUES FOR REAL-TIME MULTIFREQUENCY WAVE-FORM ANALYSIS

It is well known that, in discrete Fourier transform (DFT)-based waveform analysis of multifrequency signals, spectral parameter accuracy can be increased by windowing the time samples and interpolating the DFT coefficients. In this paper it is shown that interpolation techniques are affected very little by the number of the processed samples, so that only the characteristics of the analyzed signal and the required accuracy affect the choice of this parameter. Furthermore, the use of a polynomial fit of some relationships reduces the processing effort and allows a greater freedom in the choice of the window functions, improving both accuracy and frequency resolution requirements. © 1990 IEEE