A simplified version of the OpS algorithm for pitch stylization

In this work we present a new version of our previously published Optimal Stylization (OpS) algorithm for pitch stylization. Here we give a better perceptual representation of the pitch curve for linguistics research. While the OpS algorithm produced good stylizations for naive listeners, when deployed in a prosodic analysis tool, we observed that, under specific conditions, important details were missed in the stylized curve to an expert's ear. Changes introduced in the dynamic tonal perception model to solve these problems resulted in a simpler and more robust model. We show how the new version of the OpS algorithm is able to recover these situations while not significantly altering the original OpS curves.