How can I get from the Fast Wavelet Transform (such as using Wave++ library for example) to a Log-Frequency Spectrum? I understand fourier theory fairly well (for a computer programmer) and the basics of discrete wavelet transforms. I understand how the discrete wavelet transform can be compared to a constant-q filter bank. What I don't understand is how can the different signals from the "filter bank" output be turned into a log-frequency spectrum directly? It seems to be that each filter bank output signal (per band) has to have some sort of amplitude analysis or perhaps another transform to get the magnitude for each log-freq? if that is the case, wouldn't i need a LOT of bands if its just an amplitude calculation? Plus, the wavelet transform allows for energy near time t but perhaps not at t+k for some arbitrary log-frequency - how can looking at a log-frequency spectrum display that? the log-frequency spectrum i am needing is exactly like an FFT, except the bins are on a log-scale. I heard FWT is faster and if i can get a log-frequency spectrum from that faster than rescaling an FFT operation, then i'll do it. thanks.
How do I get from FWT to Log-Freq Spectrum?
Started by ●October 18, 2005
Reply by ●October 19, 20052005-10-19
I know what you are trying to do. You probably want to compute an octave spectrum like what B&K instrument does.I don't think there is a direct way to transform discrete wavelet transforms to FFTs. You can simply apply FFTs to multiple stages of data streams after they are decimated. Then you synthesize FFT "bins" into octave spectra. For details please contact me at DigitalSignal999[at]Yahoo.com Where [at] is an @ sign.