A new method for pitch smoothing
Wided Jlassi, Aïcha Bouzid, Noureddine Ellouze · 2016
It is possible to divine a speaker's age, gender and emotion through his pitch independently of what he is saying. However, the same words can lead to different meanings with variations intonation. Meanwhile, precise pitch estimation is more or less difficult because tracked pitch contours are not perfectly smooth curves. In this paper, we introduce a new post-processing algorithm for the tracking of pitch contours. We show the effectiveness of the proposed method using the Keele university database. Experimental results indicate that the gross pitch errors rate (GPE) is lower than comparative methods.