The behavior of a human accompanist is simulated using a hidden Markov model. The model is divided in two levels. The lower level models directly the incoming signal, without requiring analysis techniques that are prone to errors; the higher level models the performance, taking into account all the possible errors made by the musician. Alignment is performed through a decoding technique alternative to classic Viterbi decoding. A novel technique for the training is also proposed. After the performance has been aligned with the score, the information is used to compute local tempo and drive the automatic accomaniment.
An Automatic Accompanist Based on Hidden Markov Model
ORIO, NICOLA
2001
Abstract
The behavior of a human accompanist is simulated using a hidden Markov model. The model is divided in two levels. The lower level models directly the incoming signal, without requiring analysis techniques that are prone to errors; the higher level models the performance, taking into account all the possible errors made by the musician. Alignment is performed through a decoding technique alternative to classic Viterbi decoding. A novel technique for the training is also proposed. After the performance has been aligned with the score, the information is used to compute local tempo and drive the automatic accomaniment.File in questo prodotto:
Non ci sono file associati a questo prodotto.
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
