Unsupervised learning of haptic material properties.
efficient coding
haptic perception
human
materials
natural textures
neuroscience
touch
unsupervised deep learning
Journal
eLife
ISSN: 2050-084X
Titre abrégé: Elife
Pays: England
ID NLM: 101579614
Informations de publication
Date de publication:
23 02 2022
23 02 2022
Historique:
received:
13
11
2020
accepted:
09
12
2021
entrez:
23
2
2022
pubmed:
24
2
2022
medline:
18
3
2022
Statut:
epublish
Résumé
When touching the surface of an object, its spatial structure translates into a vibration on the skin. The perceptual system evolved to translate this pattern into a representation that allows to distinguish between different materials. Here, we show that perceptual haptic representation of materials emerges from efficient encoding of vibratory patterns elicited by the interaction with materials. We trained a deep neural network with unsupervised learning (Autoencoder) to reconstruct vibratory patterns elicited by human haptic exploration of different materials. The learned compressed representation (i.e., latent space) allows for classification of material categories (i.e., plastic, stone, wood, fabric, leather/wool, paper, and metal). More importantly, classification performance is higher with perceptual category labels as compared to ground truth ones, and distances between categories in the latent space resemble perceptual distances, suggesting a similar coding. Crucially, the classification performance and the similarity between the perceptual and the latent space decrease with decreasing compression level. We could further show that the temporal tuning of the emergent latent dimensions is similar to properties of human tactile receptors.
Identifiants
pubmed: 35195520
doi: 10.7554/eLife.64876
pii: 64876
pmc: PMC8865843
doi:
pii:
Types de publication
Journal Article
Research Support, Non-U.S. Gov't
Langues
eng
Sous-ensembles de citation
IM
Informations de copyright
© 2022, Metzger and Toscani.
Déclaration de conflit d'intérêts
AM, MT No competing interests declared
Références
J Neurosci. 2008 Jan 16;28(3):776-86
pubmed: 18199777
J Neurophysiol. 2017 Oct 1;118(4):2371-2377
pubmed: 28724777
Somatosens Mot Res. 2003;20(1):33-43
pubmed: 12745443
Vision Res. 2001 Nov;41(24):3145-61
pubmed: 11711140
J Neurophysiol. 2008 Mar;99(3):1422-34
pubmed: 18199814
Sci Adv. 2020 Apr 15;6(16):eaaz1158
pubmed: 32494610
Curr Biol. 2003 Mar 18;13(6):493-7
pubmed: 12646132
Elife. 2015 Dec 09;4:e10450
pubmed: 26650354
J Neurophysiol. 2015 Apr 1;113(7):3013-20
pubmed: 25744883
J Neurophysiol. 1972 Jan;35(1):122-36
pubmed: 4621505
Proc R Soc Lond B Biol Sci. 1983 Nov 22;220(1218):89-113
pubmed: 6140684
IEEE Trans Haptics. 2011 Jul-Aug;4(4):242-52
pubmed: 26963653
Proc Natl Acad Sci U S A. 2020 Mar 17;117(11):6156-6162
pubmed: 32123102
Behav Res Methods. 2019 Feb;51(1):195-203
pubmed: 30734206
Nature. 2006 Feb 23;439(7079):978-82
pubmed: 16495999
J Neurosci. 2001 Sep 15;21(18):7416-27
pubmed: 11549752
Nature. 2005 Mar 17;434(7031):387-91
pubmed: 15772663
Percept Psychophys. 2005 Jul;67(5):842-54
pubmed: 16334056
Multisens Res. 2013;26(5):429-55
pubmed: 24649528
Proc Natl Acad Sci U S A. 2013 Oct 15;110(42):17107-12
pubmed: 24082087
J Opt Soc Am A Opt Image Sci Vis. 1997 Dec;14(12):3430-1
pubmed: 9392903
Nat Hum Behav. 2021 Oct;5(10):1402-1417
pubmed: 33958744
Annu Rev Neurosci. 2001;24:1193-216
pubmed: 11520932
Curr Biol. 2011 Dec 20;21(24):R978-83
pubmed: 22192826
Proc Natl Acad Sci U S A. 2013 Jul 2;110(27):11163-8
pubmed: 23776251
Nat Neurosci. 2002 Apr;5(4):356-63
pubmed: 11896400
Nature. 1996 Jun 13;381(6583):607-9
pubmed: 8637596
J Neurophysiol. 2014 May;111(9):1792-802
pubmed: 24523522
Cogn Psychol. 1987 Jul;19(3):342-68
pubmed: 3608405
Somatosens Mot Res. 2001;18(4):253-62
pubmed: 11794728
Curr Opin Behav Sci. 2019 Dec;30:100-108
pubmed: 31886321
Nat Rev Neurosci. 2003 Jul;4(7):563-72
pubmed: 12838331
Cereb Cortex. 2020 May 14;30(5):3228-3239
pubmed: 31813989