linux - How to compare spoken audio against reference recording - language learning -
i looking way compare user submitted audio recording against reference recording comparison in order give grade or percentage language learning.
i realize un-scientific way of doing things , more gimmick anything.
my first thoughts sort of audio fingerprinting, or waveform comparison.
any ideas should looking?
this no means trivial problem solve, though there abundance of research on topic. presently successful forms of machine learning in speech recognition domain apply hidden markov model techniques.
you may want take @ existing implementations of hmm algorithms. 1 such library in stages ghmm.
perhaps better , more readily applicable problem htk.
Comments
Post a Comment