[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Perceptible information measures
We are trying to find a comparative measure of *perceptible*
information content (entropy) in sound functions, and we'd like to
avoid re-inventing the wheel. Can anyone point us to some work in
this area? The definition we're looking for would identify speech as
having high information content, but white noise as having low
information content. That is, there is a lot of spectral change in
speech signals, but almost no perceptible change in noise spectra.
One method we've thought of is to break the sound signal into frames,
calculate the loudness for each bank of a gammatone filter array,
calculate the correlations between all frame pairs, and find the
entropy of that array of values. Presumably the frames will all
correlate well for static noise (given frames of sufficient length),
but will have considerable fluctuation for more complex signals.
It would be wonderful to have some literature to refer to on this
matter, or some other ideas. Many thanks for any help!
Bret Aarden 1500 E 4th St, Dayton, OH 45403 Home: 614/361-8628
Graduate Student, School of Music, Ohio State University, Columbus, OH
Cognitive and Systematic Musicology Laboratory Lab: 614/292-7321