PresentationPDF Available

The association between statistical learning and speech perception in noise: an individual differences study

Authors:
  • Earlens Corporation

Abstract and Figures

Background & Aim: Statistical learning (SL) is a domain-general learning mechanism that is utilised by humans and primates to assess large amounts of sensory information, and extract important cues, across multiple modalities. Statistical learning has been speculated to assist with speech perception in noise. While there is evidence concerning statistical learning and its relationship to language acquisition, and vocoded speech recognition, there is little evidence concerning the role statistical learning plays in challenging listening environments. In this study, we examined the relationship of statistical learning with two speech perception in noise tasks. Methods: A total of 52 monolingual adults participated in the study. In the first study (N=22), SL performance was compared with performance of two subtests of the LiSN-S test (Cameron and Dillon, 2007). In the next experiment (N=30), SL performance was compared with another clinical speech in noise test (Kalikow, Stevens, & Elliott, 1977; Elliott, 1995) at four signal-noise ratios (SNR); 0, -5, -10, and -15 SNR. Results: The first study showed that performance on LiSN-S test was not associated with SL performance. Interestingly, SL performance was also independent of working memory and attention scores. Results from the second study showed a significant correlation between SL and speech predictability difference scores at only -10 SNR. Conclusions: These findings support the hypothesis that statistical learning is a top-down process that is employed in sufficiently challenging listening environments by normal hearing adults. The findings highlight the importance of various cognitive processes involved in speech in noise perception.
No caption available
… 
No caption available
… 
Content may be subject to copyright.
1Department of Linguistics, Macquarie University, Sydney, Australia
2The HEARing CRC
3ARC Centre of excellence for Cognition and its disorders
4Faculty of Health Sciences, University of Sydney, Sydney, Australia
Audiology Australia National Conference
25th May 2016
The association between statistical learning and
speech perception in noise: an individual
differences study
Cailyn Furze, Pragati Rao Mandikal Vasuki1, 2, 3*,
Mridula Sharma1, 2, Joanne Arciuli2, 4
Background
2
WHAT IS STATISTICAL LEARNING?
142483596342483342142596483
483596342483342142596483142
596342483342142596483142483
342483342142596483142483342
483342142596483142483342142
342142596483142483342142596
142596483142483342142596483
596483142483342142596483142
Background
3
WHAT IS STATISTICAL LEARNING?
142483596342483342142596483
483596342483342142596483142
596342483342142596483142483
342483342142596483142483342
483342142596483142483342142
342142596483142483342142596
142596483142483342142596483
596483142483342142596483142
Background
4
WHAT IS STATISTICAL LEARNING?
Implicit4
Extraction
of patterns1
1) Frost et al ., 2015
2) Perruchet et al., 2006
3) Conway & Christiansen
2005
4) Saffran et al., 1996
Different
modalities2,3
Background
5
STATISTICAL LEARNING AND OTHER COGNITIVE ABILITIES
Statistical
Learning
Digit span4
Attention4
Inhibition1,2,4
Verbal
recall1,2
Working
memory1,2,3
IQ1,2,3
1) Conway, Bauernschmidt,
Huang, & Pisoni, 2010
2) Kaufman et al. 2010
3) Siegelman and Frost 2015
4) Mandikal-Vasuki et al.
(under review)
Statistical
learning
Language
acquisition1
Word
learning1
Word
segmentation1,3
Speech
perception2
Statistical
Probability2,3
1) Saffran et al., 1996
2) Harris, 1970
3) Saffran et al., 1999
Background
STATISTICAL LEARNING AND SPEECH
Background
STATISTICAL LEARNING AND DEGRADED SPEECH
Implicit learning score
(auditory and visual)
plotted against Word
Predictability Difference
Score (auditory and audio-
visual)1
1) Conway 2010
Methods
8
STATISTICAL LEARNING and SPEECH PERCEPTION
N=35; Age- 30.4
years (SD=11)
N=30; Age- 21.3
years (SD=4.6)
Screened for normal hearing PTA, tympanometry, DPOAEs
Auditory statistical learning tested
Study 2
N=35; Age- 30.4
years (SD=11)
Study 1
Methods
9
TESTING STATISTICAL LEARNING
Behavioural task
Forced choice test between familiar triplets and
novel triplets
or
Methods
10
STATISTICAL LEARNING and SPEECH PERCEPTION
N=35; Age- 30.4
years (SD=11)
N=30; Age- 21.3
years (SD=4.6)
Screened for normal hearing PTA, tympanometry, DPOAEs
Auditory statistical learning tested
LiSN S tests
0º different voices
0º same voices
SPIN tests
0SNR
-5SNR
-10SNR
-15SNR
Study 2
N=35; Age- 30.4
years (SD=11)
Study 1
Results
11
AUDITORY STATISTICAL LEARNING and LISN-S
Low cue SRT was not associated with
performance on aSL
Results
12
STATISTICAL LEARNING and SPIN
n=23
r=0.48
p<0.05
Auditory
statistical learning
and Predictability
Difference of SPIN
lists at -10SNR
Results
13
STATISTICAL LEARNING and SPIN
0SNR
-5SNR
-10SNR
-15SNR
Conclusion
14
Statistical
learning
http://www.dispatch.com/content/stories/science/2013/12/22/1-cocktail-party-cacophony.html
creating sound valueTM 15
www.hearingcrc.org
This work was funded by the Australia Awards Scholarships and The
HEARing CRC, established and supported under the Cooperative
Research Centres Program an Australian Government Initiative. This work
was also supported by the Australian Research Council Centre of
Excellence for Cognition and its Disorders (CE110001021)
Acknowledgements
References
16
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.