Simulation and analysis of tremor in speech production

Abstract

Tremor of the muscles used for voice and speech production creates frequency and amplitude modulations in the acoustic speech signal. Frequency modulation results primarily from changes in muscle activations that alter the mass and stiffness of the vocal folds, consequently altering the fundamental frequency (F0). The primary source of amplitude modulation comes from time-varying activation of the respiratory muscles (which alter alveolar pressure), from laryngeal muscles that modify the maximum glottal area, or from interaction of voice source harmonics with the vocal tract filter. The purpose of this preliminary investigation was to analyze acoustic signals generated with an articulatory speech synthesizer in which tremor was imposed separately at each of the three anatomical sites: the respiratory system, larynx, and upper airway (pharynx). The synthesizer consisted of a voice source model of the time-varying glottal area coupled to a wave propagation model of the airways upstream and downstream of the vocal folds. This allowed for complete control of F0, alveolar pressure, glottal area, vocal tract shape, and modulation frequencies of each. Frequency and amplitude contours of the resulting acoustic signals were extracted from the waveform and analyzed and compared using customized Matlab routines. Submitted For (Speech Communication) Young Presenter Award and (Speech Communication) Best Student Paper Award.