Benjamin Tucker – benjamin.tucker@ualberta.ca
Stephanie Hedges – shedges@ualberta.ca
Department of Linguistics
University of Alberta
Edmonton, Alberta T6G 2E7
Canada
Mark Berardi – mberardi@msu.edu
Eric Hunter – ejhunter@msu.edu
Department of Communicative Sciences and Disorders
Michigan State University
East Lansing, Michigan 48824
Popular version of paper 1aSC3 (your paper version)
Presented Monday morning 11:15 AM – 12:00P PM, December 7, 2020
179th ASA Meeting, Acoustics Virtually Everywhere
Research into the perception and production of the human voice has shown that the human voice changes with age (e.g., Harnsberger et al., 2008). Most of the previous studies have investigated speech changes over time using groups of people of different ages, while a few studies have tracked how an individual speaker’s voice changes over time. The present study investigates three male speakers and how their voices change over the last 30 to 50 years of their lives.
We used publicly available archives of speeches given to large audiences on a semi-regular basis (generally with a couple of years between each instance). The group of speeches was given during the last 30-50 years of each speaker’s life, meaning that we have samples ranging from the speakers’ late 40s to early 90s. We extracted 5-minute samples (recordings and transcripts) from each speech. We then used the Penn forced-alignment system (this system finds and marks the boundaries of individual speech sounds) to identify word and sound boundaries. Acoustic characteristics of the speech were extracted from the speech signal using a custom script using the Praat software package.
In the present analysis, we investigate changes in the vowel space (the acoustic range of vowels a speaker has produced), fundamental frequency (what a listener hears as pitch), the duration of words and sounds (segments), and speech rate. We model the acoustic characteristics of our speakers using Generalized Additive Models (Hastie & Tibshirani, 1990), which allows for an investigation of non-linear changes over time.
The results are discussed in terms of vocal changes over the lifespan in the speakers’ later-years. Figure 1 illustrates the change in one speaker’s vowel space as he ages. We find that for this speaker the vowel space shifts to lower frequencies as he ages.
Figure 1 – An animation of Speaker 1’s vowel space and how it changes over a period of 50 years. Each colored circle represents a different decade.
We also find a similar effect for fundamental frequency across all three speakers, Figure 2, where the average fundamental frequency of their voices gets lower and lower as they age and then starts to get higher after the age of 70. This effect is the same for word and segment duration. We find that on average as our three speakers age their speech (at least when giving public speeches) gets faster and then slows down after around the age of 70.
Figure 2: Average fundamental frequency of our speakers’ speech as they age.
Figure 3: Average speech rate in syllables per second of our speakers’ speech as they age.
While on average our three speakers show a change in the progression of their speech at the age of 70, each speaker has their own unique speech trajectory. From a physiological standpoint, our data suggest that with age come not only laryngeal changes (changes to the voice) but also a decrease in respiratory health – especially expiratory volume – as has been reflected in previous studies.