Content area
Abstract
Electrolarynx users frequently report that the device produces a very unpleasant robotic voice, stating the speech generated by these devices is monotonous and mechanical. Methodology was perfected to extract the natural glottal volume-velocity waveforms from normal subjects. Then three experiments were designed to identify the parameters of the electrolarynx source excitation pulse that account for these negative reactions. In Experiment 1 the volume-velocity waveforms from five different electrolarynx devices were measured. The natural and electrolarynx source waveform data was then used to construct a computerized model of an electrolarynx with a modified version of the Klatt formant synthesizer. Synthesizing an electrolarynx source pulse in software was done so that electrolarynx-like vowel or sentence stimuli could be generated more rapidly than the hardware construction of a new electrolarynx each time a new excitation pulse was desired. In the final two experiments, synthetic electrolarynx CVC syllable stimuli were presented to listeners. The pulse parameters were manipulated and listener preferences were measured. The results of this work identified how these parameters could be modified, for incorporation into electrolarynx hardware with the ultimate goal of improving naturalness.
