From silk to sand: Multiple dimensions of perceived softness

Data related to the publication Manuscript: From silk to sand: Multiple dimensions of perceived softness Perceiving object softness is a crucial ability in many aspects of life, from choosing an avocado to picking your clothes. Traditionally, haptic softness has been equated with the compliance of elastic materials or springs, however, people use the attribute ‘soft’ to describe a wide variety of materials that differ substantially in their physical properties. For example, both, silk and sand are soft, but in very different ways. In this study 32 participants explored soft and non-soft materials haptically and rated them on several attributes. Using the semantic differential technique and video analysis, we show here that haptic softness has, in fact, four dimensions and that each softness dimension is associated with a distinct set of hand and finger movements during active exploration. Taken together our findings suggest that judging the softness of materials is a more heterogeneous process than originally thought, also affecting how a material is explored. Below script for classification from Mathworks Inc. uses built-in Matlab functions, we only changed variable names: clear all, close all, clc<br> load(data)<br> for i = 1:length(obslist) <br> index(i) = find([mater_names{:,1}] == obslist(i,1)); <br> end for c = 1:length(obslist)<br> idx = c;%obslist(c,1);<br> mater_categ(c,1) = mater_names(index(idx),3);<br> mater_categ(c,2) = mater_names(index(idx),2);<br> end<br> unique(mater_categ) X = epm; % EP frequencies are measurements<br> Y = mater_categ(:,1); % softness dimensions t = templateSVM('Standardize',1,'KernelFunction','gaussian'); Mdl = fitcecoc(X,Y,'Learners',t,'FitPosterior',1,...<br> 'ClassNames',{'deformable';'fluid';'granular';'rough';'surface'},...<br> 'Verbose',2);<br> [label,~,~,Posterior] = resubPredict(Mdl,'Verbose',2);<br> Mdl.BinaryLoss<br> idx = randsample(size(X,1),10,1);<br> % Mdl.ClassNames<br> table(Y(idx),label(idx),Posterior(idx,:),...<br> 'VariableNames',{'TrueLabel','PredLabel','Posterior'}) t = templateSVM('Standardize',1)<br> Mdl = fitcecoc(X,Y,'Learners',t,...<br> 'ClassNames',{'deformable';'fluid';'granular';'rough';'surface'});<br> CVMdl = crossval(Mdl,'Leaveout','on');<br> oosLoss = kfoldLoss(CVMdl)