基于KNN SVM NN TREE NB等多种不同分类器评估的自动面部表情识别系统附matlab代码.zip

function expl = shapePCA(X,whichPCs,numSDs,animatedPlot,MDSorGPA,PAtype) %SHAPEPCA Analyse shape variation using principal component analysis. % SHAPEPCA(X) displays plots for the variation at +/- 3 SDs along % the first 4 PCs. % % Z = SHAPEPCA(X) is the percentage variation explained by each principal % component. % % There are several optional arguments (the defaults shown in square % brackets): % whichPCs [1:4]: the PCs for which to plot; % numSDs [3]: number of SDs for which to show variation along each PC; % animatedPlot [false]: if true then an animated plot is shown for % the variation along PC whichPCs(1); % MDSorGPA ['mds']: if false or 'mds' then MDS is used to calculate % the mean shape, else if true or 'gpa' then GPA is used. % Alternatively MDSorGPA can be a user-prescribed m-by-k % configuration; % PAtype [false]: if false then partial Procrustes analysis is used; % if true then full Procrustes analysis is used. % Written by Simon Preston (http://www.maths.nott.ac.uk/~spp), 2008 if nargin < 6 || isempty(PAtype), PAtype = false; end; if nargin < 5 || isempty(MDSorGPA), MDSorGPA = 'mds'; end; if nargin < 4 || isempty(animatedPlot), animatedPlot = false; end; if nargin < 3 || isempty(numSDs), numSDs = 3; end; if nargin < 2 || isempty(whichPCs), whichPCs = 1:4; end; [m,k,n] = size(X); Y = unitFrobNorm(centre(X)); if (numel(MDSorGPA)==1 && (MDSorGPA==false)) || strcmpi(MDSorGPA,'mds') Ymean = MDSmean(Y,true); Yp = OPA(Ymean,Y,PAtype); elseif (numel(MDSorGPA)==1 && MDSorGPA==true) || strcmpi(MDSorGPA,'gpa') Yp = GPA(Y,PAtype); Ymean = mean(Yp,3); elseif size(MDSorGPA) == size(X(:,:,1)) Ymean = MDSorGPA; Yp = OPA(Ymean,Y,PAtype); else error('Fifth argument (MDSorGPA) is invalid: consult ''help shapePCA'''); end % Project onto tangent space YpT = zeros(m*(k),n); projector = eye(m*(k)) - vec(Ymean)*vec(Ymean)'; for i = 1:n YpT(:,i) = projector*vec(Yp(:,:,i)); end covar = cov(YpT'); [evecs, variances] = eigSort(covar); expl = 100*(variances/sum(variances)); YPCplus = zeros([m k m*(k)]); YPCminus = zeros([m k m*(k)]); for i = 1:(m*(k)) v = evecs(:,i)*numSDs*sqrt(variances(i,i)); YPCplus(:,:,i) = reshape(v + sqrt(1-v'*v)*vec(Ymean),[m k]); YPCminus(:,:,i) = reshape(-v + sqrt(1-v'*v)*vec(Ymean),[m k]); end if animatedPlot if numel(whichPCs) > 1 warning(['When animatedPlot==true, only a plot for the first ',... 'specified PC [i.e. whichPCs(1)] is shown']); end figTitle = ['PC ',num2str(whichPCs(1)),'. Explains ',... num2str(expl(whichPCs(1)),3),... '% of variation', ... ' (plotted along ',num2str(numSDs), ' SDs)']; plotShapeAnimated(... YPCplus(:,:,whichPCs(1)),YPCminus(:,:,whichPCs(1)),figTitle); else h = zeros(1,length(whichPCs)); figh = zeros(1,length(whichPCs)); for iPC = 1:length(whichPCs) PC = whichPCs(iPC); figh(iPC)=figure; axis vis3d; axis equal; hold on; h(iPC) = gca; figTitle = ['PC ',num2str(PC),'. Explains ',... num2str(expl(PC),3),... '% of variation', ... ' (plotted along ',num2str(numSDs), ' SDs)']; set(gcf,'Name',figTitle); % title(figTitle) if m == 2 for i = 1:k plot(... cat(1,YPCplus(1,i,PC),Ymean(1,i)),... cat(1,YPCplus(2,i,PC),Ymean(2,i))); plot(... cat(1,Ymean(1,i),YPCminus(1,i,PC)),... cat(1,Ymean(2,i),YPCminus(2,i,PC)),'Color',[.9 .9 1]) plot(Ymean(1,i),Ymean(2,i),'.'); end elseif m == 3 for i = 1:k plot3(... cat(1,YPCplus(1,i,PC),Ymean(1,i)),... cat(1,YPCplus(2,i,PC),Ymean(2,i)),... cat(1,YPCplus(3,i,PC),Ymean(3,i))); plot3(... cat(1,Ymean(1,i),YPCminus(1,i,PC)),... cat(1,Ymean(2,i),YPCminus(2,i,PC)),... cat(1,Ymean(3,i),YPCminus(3,i,PC)),'Color',[.9 .9 1]) plot3(Ymean(1,i),Ymean(2,i),Ymean(3,i),'.'); end end end % Links the axes properties hlink = linkprop(h,{'CameraPosition','CameraUpVector',... 'CameraViewAngle'}); key = 'graphics_linkprop'; setappdata(h(1),key,hlink); set(h(1),'CameraViewAngle',9); % Zoom out a little tilefigs(figh); end %plotFaceAnimated(PC_lines_Y(:,:,PC),PC_lines_Yb(:,:,PC));