سؤال

I make a emotion-recognition program using comirva package.

and I have a lot of wav files about emotion like angry, happy..

to make a gmm, first, i extracted mfcc features from one of argry files.

what I confused is that how to change mffccs to pointlist.

wich one is correct? i) changing mfcc extracted from one file to one pointList ii) chaiging mfccs extracted from many files to one pointlist

هل كانت مفيدة؟

المحلول

correct is

iii) chaiging mfccs extracted from many files which belong to same emotion (e.g. angry) to one pointlist and to train GMM on them. As a result you should get a set of GMMs which correspond to emotions.

This algorithm for emotion classification is sort of senseless. To detect emotion one requires some other features like pitch and so on. I recommend you to look on real emotion recognition libraries first like openart/openear:

https://sourceforge.net/projects/openart/

مرخصة بموجب: CC-BY-SA مع الإسناد
لا تنتمي إلى StackOverflow
scroll top