A strategy for LF-based glottal-source & vocal-tract estimation on stationary modal singing

This paper presents a methodology for estimation and modeling of the glottal source and vocal-tract information. The strategy proposes a simplified framework based on the characteristics of stationary singing following a selection of glottal pulse model candidates driven by a single shape parameter. True-Envelope based models are applied, allowing efficient modeling of the observed filter information and accurate cancellation of the glottal source contribution in the spectrum. According to experimental studies on synthetic and real signals the methodology observes adequate approximation of the source and filter information, leading to natural resynthesis quality using synthetic glottal excitation. The proposed estimation framework represents a promising technique for voice transformation on stationary modal voice.