Voice jitter estimation using high-order synchrosqueezing operators

MIRAMONT, JUAN M.; COLOMINAS, MARCELO ALEJANDRO; SCHLOTTHAUER, GASTON

IEEE/ACM Transactions on Audio, Speech, and Language Processing

IEEE

Año: 2020 p. 1 - 1

2329-9290

Voice jitter is defined as a random perturbationof the glottal cycle duration which can be useful for voiceparametrization and that usually depends on finding fiducialpoints in this signal. In this paper, a novel application of theFourier-based high-order synchrosqueezing (FSSTN) operatorson voice signals is introduced for voice jitter estimation withoutperiod-segmentation. To this end, an innovative interpretationof the relative jitter formula in terms of the total variation ofthe sequence of periods is proposed. This allows us to derivean algorithm for jitter estimation that uses the (continuous)instantaneous fundamental frequency of the signal and its firstderivative (chirp rate) which can be obtained, respectively, fromthe local complex frequency and the local complex modulationFSSTN operators. Numerical experiments using synthetic signalswith known true jitter show that this novel approach yieldssimilar results to other state-of-the-art method, PRAAT, for truejitter within the range [0.2%, 1.2%], and that it outperformsPRAAT for true jitter values in the range [1%, 15%]. Thehere proposed method seems a promising tool for voice jitterestimation and constitutes a novel application of the highorder synchrosqueezing operators for voice signals with potentialimpact on jitter modeling and on the clinical field.