Evaluation of the pitch estimation algorithms in the monopitch and multipich cases
Résumé
Reliably tracking the fundamental frequency F0 of the components is an important step in the separation of
superimposed speech signals. Several Pitch Estimation Algorithms (PEAs) are potentially usable and a rigorous
evaluation method is needed. However, even in the monopitch case, many variations between them render such a
comparison difficult. The F0min-F0max interval extent, the use of a priori information on the whole sequence or
database and above all the arbitrary voicing threshold setting lead to large differences in the results. These biases
can be removed by setting the F0 bounds to fixed values acceptable for many voices, by proceeding with the
evaluation on a strictly frame-to-frame basis, and by fixing the voicing threshold in order to get an equal error
rate for overvoiced and undervoiced frames. In the multipitch case any frame may exhibit 0, 1 or 2 valid voicing
according to the coincidence between the voiced and unvoiced parts of both signals. This problem is treated by
defining a metric linking the PEAS’s hypotheses to the pitch values of the isolated signals. The proposed
methodology is applied to several PEAs on several databases in the monopitch case.
Origine : Fichiers produits par l'(les) auteur(s)