Spectral analysis of the human voice: a potentially useful tool in rehabilitation

Abstract

Aim: The study was aimed at providing normal data from the spectral analysis of the human voice. This step of the research has a bearing in the field of speech and language retraining. The possibility of providing quantitative data would hasten monitoring of possible improvements along a rehabilitation program.

Methods: A cross-sectional, observational design was applied. Thirty-two men, 30 women and 34 children were the healthy volunteers participating to the study. The adult participants had to repeat the 12 words of the ENPA instrument for Italian language. The children repeated words from the Fanzago test. Real Time Pitch Model 5121 manufactured by Kay PENTAX recorded the voice. The following spectral descriptors were obtained for each word: mean frequency, standard deviation, loudness, duration, jitter and shimmer. Test-retest performance was also checked.

Results: As expected, women had a higher mean frequency than men (119+/-3 v. 193+/-5 Hz), but the variation coefficient (CV: SD/mean) was the same, namely 0.1. By contrast, children had a significantly higher mean frequency (245+/-35 Hz) and a lower CV. Under this respect, groups could be statistically divided. Loudness was practically the same among the groups, whereas the children had short time duration in word uttering. Children had also the highest jitter values, whereas shimmer values were higher in men.

Conclusions: Spectral analysis of the human voice is recommended in each laboratory of speech and language rehabilitation. The usual testing is not replaced but can exploit the accurate definition of voice descriptors.