Glottal source analysis of emotional speech

Emotional speech makes speech more expressive, emotional speech conversion is needed in many systerms. Analyzing of glottal source wave plays an important role in emotional speech conversion. The purpose of this paper is to analyze glottal source of emotional speech for emotional speech conversion based on the Auto-Regressive eXogenous (ARX) model combined with Liljencrant-fant (LF) model, in which the Glottal Clsure Instant (GCI) and Glottal Opening Instant (GOI) are two important parameters and greatly affect the accuracy of the ARX-LF model. Therefore, a mean-based signal method is suggested to improve the estimation accuracy of GCI, and GOI is estimated from the Hilbert envelope of LP residual. The ARXLF model with accurate GCI and GOI is applied for analysis of glottal source of emotional speech. The results show that the proposed approach improve the accuracy of glottal source wave of speech, and the different glottal source waves of different emotional speech can be obtained