Low-Frequency Energy Deficit in Electrolaryngeal Speech

Abstract

The present exploratory project was undertaken (a) to determine the relative strength of low-frequency energy in the output of one widely used electronic artificial larynx (Servox) and (b) to assess the relative strength of low-frequency energy in vowels produced by users of this type of artificial larynx. We hypothesized that the outputs of electronic artificial larynges and the vowels produced by laryngectomized users of these devices would be characterized by significant deficits in low-frequency energy level. Five users of electronic larynges and 10 normal speakers (5 female and 5 male) provided the speech samples. Results of spectral analyses indicated that there was a significant deficit in low-frequency energy both in the acoustic signals generated by a Servox electronic larynx and in vowels produced by laryngectomized users of this type of electronic larynx. Based on these findings, a second order filter was designed and implemented digitally to compensate for the observed deficit in low-frequency energy. A perceptual experiment was completed to evaluate the effect of low-frequency enhancement on perceived speech quality. Ninety-eight percent (+/- 2%) of the responses of listeners indicated that low-frequency enhanced speech samples had better vocal quality or were more pleasant to listen to than the original speech samples. We conclude that consideration for enhancing low-frequency characteristics is warranted in the design of improved prosthetic devices for alaryngeal speakers.