引用LSVT数据集的论文1
需积分: 0 41 浏览量
2022-08-03
19:13:43
上传
评论
收藏 924KB PDF 举报
IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 22, NO. 1, JANUARY 2014 18
1
Objective Automatic Assessment of Rehabilitative
Speech Treatment in Parkinson’s Disease
Athanasios Tsanas, Max A. Little, Cynthia Fox, and Lorraine O. Ramig
Abstract—Vocal performance degradation is a common
symptom for the vast majority of Parkinson’s disease (P
D)
subjects, who typically follow personalized one-to-one periodic
rehabilitation meetings with speech experts over a long-term
period. Recently, a novel computer program cal
led Lee Silverman
voice treatment (LSVT) Companion was developed to allow PD
subjects to independently progress through a rehabilitative treat-
ment session. This study is p a rt of the asses
smentoftheLSVT
Companion, aiming to investigate the potential of using sustained
vowel phonations towards objectively and automatically repli-
cating the speech experts’ assessments o
f PD subjects’ voices as
“acceptable” (a clinician would allow persisting during in-person
rehabilitation treatment) or “unacceptable” (a clinician would not
allow persisting during in-person
rehabilitation treatment). We
characterize each of the 156 sustained vowel /a/ phonations with
309 dysphonia measures, select a parsimonious subset using a
robust feature s election algori
thm, and automatically distinguish
the tw o cohorts (acceptable versus unacceptable) with about
90% overall accuracy. Moreover, we illustrate the potential of
the proposed methodology as a p
robabilistic decision support
tool to speech experts to assess a phonation as “acceptable” or
“unacceptable.” We envisage the findings of this study being a
first step towards improvi
ng the effectiveness of an automated
rehabilitative speech assessment tool.
Index Terms—Decision support tool, Lee Silverman voice treat-
ment (LSVT), Parkinson’s disease (PD), speech rehabilitation.
I. INTRODUCTIO N
P
ARKINSON’S disease (PD) is a chronic neurodegenera-
tive disorder characterized by the progressive deteriora-
tion of motor function and the emergence of considerable non-
Manuscript received June 25, 2013; revised October 19, 2013; accepted
November 24, 2013. Date o f publication December 05, 2013; date of current
version January 06, 2014. This work was supported in part by the National
Institutes of H ealth (NIH) un der Grant 1R43DC0109 56-01 . The work of M.
A. Little was suppo rted by the Wellcome Trust under Grant WT090 651MF.
The work of A. Tsanas was supported in part by the Engineering and Physical
Sciences Research Counc il (EPSRC), U.K., and in part by the Wellcome Trust
through a Centre Grant 098461/Z/12/Z, “The University of Oxford Sleep and
Circadian Neuroscience Institute (SCNi ).”
This paper has su pplementary down loadable material ava ilab le at http://iee-
explore.ieee.org, provided by the authors. This includes an Excel file which is
34 Kb in size.
A. Tsanas is with the Institu te of Biom edical Engineering, Department of
Engineering Science, University of Oxford, OX1 3PJ Oxford, U.K., and with the
Wolf son Centre for Mathematical Biology, Mathematical Institute, University
of Oxford, OX1 3LB Oxford, U.K. (e-mail: tsanas@maths.ox.ac.uk).
M. A. Little is with the Medi a Lab, Massachusetts Institute of Technology,
Cambridge, MA 02139 USA (e-mail: maxl@mit.edu).
C. Fox and L. O. Ramig are with the Speech, Language, and Hea ring Science,
University of Colorado, Boulder, CO 80309 USA, and with the National Center
for Voice and Speech, Denver, CO 80204 USA (e-mail: cynthia.fox@lsvtglobal.
com; lorraine.ramig@colorado.edu)
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Obje ct Iden t ifier 10.1109/TNSRE.2013.2293575
motor problems [1]. It is estimated there are at least 100 PD sub-
jects per 100 000 in the po pulatio n [2], and some studies ha ve
suggested PD prevalence m ay be underestimated [3]. Vocal im-
pairment is reported in the vast majority of PD subjects, and
approximately 2 9% of those con sider it one of their greatest hin-
drances associated with the disease [4]. Moreover, speech per-
formance degradation may be amongst the first sympto m s of PD
onset [5].
Typical vocal impairment symptoms include r educed loud-
ness, mon oto ne, hoarseness, b reathiness (noise), im precise ar-
ticulation, and vocal tremor [6]. The extent of vocal im pairment
can be assessed using sustained vowel phonations, or running
speech. It can be argued that while some of the vocal deficien-
cies in running speech caused by PD (e.g., sequences of conso-
nants and vowels) may not occur in sustained vowels, the anal-
ysis of running speech is inherently more complex due to ar-
ticulatory and o ther linguistic confounds [7], [8]. Consequently,
sustained vowel phonations, where the speaker attempts to pro-
duce a vowel sound as steady as possible (in terms of amplitude
and frequency) and for as long as possible, are commonly used
in clinical practice [8]. Both clinical practice and extensive re-
search have shown that the sustained vowel “ahh
” (denoted
/a/) may be sufficient for many voice assessment applications
[8]–[11], and for PD voice assessment in p articular [12]–[17].
Clinical speech signal processing algorithms ( algorithm ic
tools extracting clinically useful information from speech
signals) are collectively known as dyspho nia m easures.From
the point of view of speech experts and ear, nose, and throat
surgeons, there is an important distinction between voice (the
sound produced by the larynx) and speech, which is the in-
tegrated process of v oice and articulation. Despite the subtle
difference in the narrow definition of voice and speech, I. Titze
asserts “in the b roader sense voice is synonymous with speech”
[8]. Strictly speaking, the dysphonia measures discussed in this
study are entirely based on voice rather than speech since we
do not attempt to characterize articulation problems.
LSVT LOUD is a standardized, research-based speech treat-
ment proto col with established efficacy for PD [18]–[20]. A sig-
nificant challenge is how to scale accessibility to this speech
treatment program that requires a sustained, intensive treatment
regime when ther e are not enough clinician s to del iver all t he
treatment that is needed. Advances in computer and w eb-based
technology offer solutions to the problems of treatment acces-
sibility, efficacious treatment delivery, and long-term mainte-
nance in rehabilitation [21]. Such tech nology may alleviate the
barrier of inadequate numbers of clinicians to deliver in-person
therapy, enhance the feasibility of delivering intensive treatment
requirements, and relieve the logistical burden of traveling to
1534-4320 © 2013 IE EE. P e rso nal use is permitted, but republication /redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
评论0