Abstract : We present a system for recognition and comparison of acoustic signals of singing voices, aiming at the sensorimotor learning of Byzantine music. Οur system is trained with chants performed by the expert and it recognizes any of the interpretations (expert/learner) from a small corpus. Τhe system can evaluate in real time the distance of the performances. The results are satisfying, regarding (a) the training of expert model with the fundamental frequency (pitch) F0 and the MFCC coefficients showing in parallel the optimum combination for training (b) regarding the recognition of the trainee’s singing voice, c) the real-time estimation of the distance between expert’s and learner’s performance.
https://hal-mines-paristech.archives-ouvertes.fr/hal-01508636
Contributeur : Sotiris Manitsaris
<>
Soumis le : vendredi 14 avril 2017 - 13:29:46
Dernière modification le : lundi 12 novembre 2018 - 10:55:24
