I've update the mfcc calculation to calculate the pause at the begining
and ending of tracks. The lead in and lead out of the tracks is then stripped
before the mfccs are calculated.
This should improve the quality of lists generated by the markov chainer, as
it won't be attempting to match silences between tracks on the playlist
anymore.
Of course, this costs a few more days of CPU time to calculate MFCCs.
jack |
posted Mon Sep 19 15:10:44 2005 |
#
category:
projects/markov
A weblog by Jack Cummings