Arshia Cont, Diemo Schwartz and, Norbert Schnell
This paper describes our attempt to make the Hidden Markov Model (HMM) score following system developed at IRCAM sensible to past experiences in order to adapt itself to a certain style of performance of musicians on a particular piece. We focus mostly on the aspects of the implemented machine learning technic pertaining to the style of performance of the score follower. To this end, a new observation modeling based on Gaussian Mixture Models is developed which is trainable using a novel learning algorithm we would call automatic discriminative training. The novelty of this system lies in the fact that this method, unlike classical methods for HMM training, is not concerned with modeling the music signal but with correctly choosing the sequence of music events that was performed.