Voiced Excitations

摘要

To more easily obtain a voiced excitation function for speech characterization, measurements of skin motion, tracheal tube, and vocal fold, motions were made and compared to EM sensor-glottal derived excitation functions during voiced speech. Skin motions were measured at 9 locations on the neck and face. Typically 2 micron amplitudes were observed depending upon the sound and location. The location below the vocal folds provided the most robust information. Tracheal wall motions above and below the glottis were measured with a low power EM interferometer attached to a narrow beam antenna. Motions of 5-10 microns were measured with best results by using the E field polarized in the vertical direction at the subglottal location. Tracheal motion and vocal fold signals are quite different in amplitude, in spectral response, and in E-field polarization response, from each other. Both skin and trachea signals showed spectral detail up to about 500 and 800 Hz respectively. Methods to use these data to generate sufficiently good excitation functions of voiced speech are discussed.