Usually, wavelet transform is real, so the output of a real sine is
still a real sine. While that in STFT, the transform is complex, so
the spectrogram is corresponding to the square of amplitude. For
example, s(t) = A sin(xt), the intensity of frequency x at certain
time instant in spectrogram is A^2, while that associated with
scalogram is proportional to [A sin(xt)]^2. The other difference is
the scale vs frequency. If only dyadic wavelet transform is performed,
the frequency band is divided as [fs/4 fs/2], [fs/8 fs/4], [fs/16
fs/8]... whereas the frequency is sampled evenly in STFT. So if you
want to get a even frequency sampling in scalogram, you can't use the
dyadic sampling.
NI will provide an analytic (complex, continuous) wavelt transform in
next wavelet toolkit release. It can perform the traditional scalogram
and the STFT comparable scalogram.

NISH DSP

Reply via email to