Partners:
Everis, Spain
ETH, Switzerland
UZH, Switzerland
Freiburg, Germany
MA Systems, UK
Bristol, UK
Xiwrite, Italy
Ultrasis, UK
Jaume, Spain
Valencia, Spain
Lanzhou, China
EU-Grant (FP7):
248544
|
Normative Studies of 192 Healthy Subjects
Prior to using the speech analysis method as a standard tool for the assessment
of the affective state of a speaker, several questions have to be addressed:
(1) the optimum recording time containing enough information for a reliable
quantification of speech parameters; (2) the distribution of the inter-individual
scattering of speech parameters in the general population; (3) the intra-individual
stability of speech parameters over time; (4) the differences between dialect and
non-dialect, and between affect-neutral and affect-charged speech; and (5) the
amount of variance explainable by the external factors age, sex and education.
Learning to Recognize
To address these questions, we have carried out a normative study with 192 healthy
volunteers, stratified according to age, sex and education. The specific design of
this study with 3 different types of text and 2 repeated measurements at 14 day
intervals allowed us to analyze the intra-individual variation of speech parameters
over time, their sensitivity to form and content of spoken text, and the
inter-individual scattering of speech parameters. Thus, we were able to derive
normative values of the general population and to learn to distinguish between
"natural" fluctuations and "significant" changes which may encompass short-term
reactions to the immediate environment or longer persisting deviations from
"normality". Natural fluctuations are defined through the interval of
plusminus two standard deviations around mean values(90%). Deviations
beyond these thresholds are regarded as significant
Language-Dependence of Speech Parameters
As speech parameters are language-dependent, OPTIMI relies on the normative study
previously carried out at Zurich (Switzerland: German), along with a study of 120
healthy volunteers carried out at Bristol (UK: English), and a study of 120 healthy
volunteers carried out at Valencia (Spain: Spanish). The test persons are ascertained
on the basis of the Zurich Health Questionnaire (ZGF) and invited twice at 14-day
intervals to the recording studio where they are asked to present 3 different types
of text: (1) automatic speech, (2) reading out loud an emotionally neutral text, and
(3) reading out loud an emotionally stimulating text. The speech recordings
are carried out according to the following scheme:
-
Counting out loud from 1 to 40 (0.5 min)
-
Short pause (0.5 min)
-
Emotionally neutral text from a children's book (2 min)
-
Short pause (0.5 min)
-
Emotionally stimulating text from a well-known author (2 min)
-
Counting out loud from 1 to 40 (0.5 min)
|
|
Stability of the speech parameters "mean vocal pitch" and "6db-bandwidth"
over time among 91 healthy volunteers: the first assessment is plotted along the
x-axis and the second assessment 14 days later along the y-axis.
While mean vocal pitch displays a high stability over time, the speech parameter
"6db-bandwidth" (measuring intonation) is much less stable. The experimental condition
is "counting" and "freehand speech".
|