Sound: To MelSpectrogram...


A command that creates a MelSpectrogram object from every selected Sound object by band filtering in the frequency domain with a set of triangular filters.
The filter functions used are all triangular in shape on a mel frequency scale. Each filter function depends on three parameters, the lower frequency f_{l}, the central frequency f_{c} and the higher frequency f_{h}. On a mel scale, the distances f_{c}f_{l} and f_{h}f_{c} are equal for each filter. The filter function is as follows:
H(f) = 0 for f ≤ f_{l} and f ≥ f_{h} 
H(f) = (f  f_{l}) / (f_{c}  f_{l}) for f_{l} ≤ f ≤ f_{c} 
H(f) = (f_{h}  f) / (f_{h}  f_{c}) for f_{c} ≤ f ≤ f_{h} 
In general the number of filter values stored in each frame of the MelSpectrogram is an order of magnitude smaller than the number of sound samples in the corresponding analysis frame.
Links to this page
© djmw, October 22, 2014