Aligning music audio with symbolic scores using a hybrid graphical model