TY - JOUR
T1 - Frequency specificity of amplitude envelope patterns in noise-vocoded speech
AU - Ueda, Kazuo
AU - Araki, Tomoya
AU - Nakajima, Yoshitaka
N1 - Funding Information:
Thanks are due to Gerard B. Remijn for providing valuable suggestions on a draft of this article. Funding: This work was supported by the Japan Society for the Promotion of Science [grant numbers 19103003 , 25242002 , 20330152 , 17K18705 ]; the Ministry of Education, Culture, Sports, Science and Technology, Japan [Grant-in-Aid for the 21st Century COE program].
Publisher Copyright:
© 2018 The Authors
Copyright:
Copyright 2019 Elsevier B.V., All rights reserved.
PY - 2018/9
Y1 - 2018/9
N2 - We examined the frequency specificity of amplitude envelope patterns in 4 frequency bands, which universally appeared through factor analyses applied to power fluctuations of critical-band filtered speech sounds in 8 different languages/dialects [Ueda and Nakajima (2017). Sci. Rep., 7 (42468)]. A series of 3 perceptual experiments with noise-vocoded speech of Japanese sentences was conducted. Nearly perfect (92–94%) mora recognition was achieved, without any extensive training, in a control condition in which 4-band noise-vocoded speech was employed (Experiments 1–3). Blending amplitude envelope patterns of the frequency bands, which resulted in reducing the number of amplitude envelope patterns while keeping the average spectral levels unchanged, revealed a clear deteriorating effect on intelligibility (Experiment 1). Exchanging amplitude envelope patterns brought generally detrimental effects on intelligibility, especially when involving the 2 lowest bands (≲1850 Hz; Experiment 2). Exchanging spectral levels averaged in time had a small but significant deteriorating effect on intelligibility in a few conditions (Experiment 3). Frequency specificity in low-frequency-band envelope patterns thus turned out to be conspicuous in speech perception.
AB - We examined the frequency specificity of amplitude envelope patterns in 4 frequency bands, which universally appeared through factor analyses applied to power fluctuations of critical-band filtered speech sounds in 8 different languages/dialects [Ueda and Nakajima (2017). Sci. Rep., 7 (42468)]. A series of 3 perceptual experiments with noise-vocoded speech of Japanese sentences was conducted. Nearly perfect (92–94%) mora recognition was achieved, without any extensive training, in a control condition in which 4-band noise-vocoded speech was employed (Experiments 1–3). Blending amplitude envelope patterns of the frequency bands, which resulted in reducing the number of amplitude envelope patterns while keeping the average spectral levels unchanged, revealed a clear deteriorating effect on intelligibility (Experiment 1). Exchanging amplitude envelope patterns brought generally detrimental effects on intelligibility, especially when involving the 2 lowest bands (≲1850 Hz; Experiment 2). Exchanging spectral levels averaged in time had a small but significant deteriorating effect on intelligibility in a few conditions (Experiment 3). Frequency specificity in low-frequency-band envelope patterns thus turned out to be conspicuous in speech perception.
UR - http://www.scopus.com/inward/record.url?scp=85048790672&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85048790672&partnerID=8YFLogxK
U2 - 10.1016/j.heares.2018.06.005
DO - 10.1016/j.heares.2018.06.005
M3 - Article
C2 - 29929750
AN - SCOPUS:85048790672
SN - 0378-5955
VL - 367
SP - 169
EP - 181
JO - Hearing Research
JF - Hearing Research
ER -