Extended Data Fig. 9: Surface texture classification results using SS values recorded after mechanical stimulation. | Nature Materials

Extended Data Fig. 9: Surface texture classification results using SS values recorded after mechanical stimulation.

From: Bio-inspired artificial mechanoreceptors with built-in synaptic functions for intelligent tactile skin

Extended Data Fig. 9

For ML, we collected 2560 EPSC signals (1280 SA signals and 1280 FA signals) using the ASMRs array, spanning 16 categories of textures. a-c, Confusion matrices illustrating the surface texture classification accuracy (highest) using features from (a) SA, (b) FA ASMRs, and (c) multimodal fusion of SA and FA (SA/FA) features for the classification of 16 texture categories (that is, 1: Cotton 100%, 2: polyester 100%, 3: rayon 100%, 4: nylon 100%, 5: acetate 100%, 6: cotton 65%-polyester 35%, 7: polyester 85%-spandex 15%, 8: cotton 52%-nylon 48%, 9: polyester 56%-rayon 44%, 10: nylon 54%-rayon 46%, 11: cashmere 100%, 12: linen 100%, 13: cowhide 100%, 14: skin, 15: nitrile 100%, and 16: Al foil). d-f, Receiver operating characteristic (ROC) curves for texture classification using (d) SA features, (e) FA features, and (f) combined SA/FA features.

Source data

Back to article page