Automatic Temporal Alignment for Pitch Estimation Evaluation

Abstract

Pitch estimation plays a critical role in various fields from speech analysis and synthesis to medical diagnosis. Despite the importance of pitch estimation, the evaluation of pitch estimation methods often yields inconsistent results and unfair comparisons. This inconsistency largely stems from the lack of temporal alignment between the ground-truth sequence and pitch estimation sequence. To address this issue, this paper proposed an automatic approach to temporally align pitch sequences by determining an optimal target offset. Experiments were conducted on mainstream pitch estimation methods and datasets, and results showed effective improvement in evaluation accuracy. A MATLAB implementation of the proposed method is publicly available online.