bipartitematchingdemo.m~

利用HMM的方法的三种语音识别算法

% Consider matching sources to detections

%  s1 d2  
%         s2 d3
%  d1

%a  = optimalMatching([52;0.01])

% sources(:,i) = [x y] coords
sources = [0.1 0.7; 0.6 0.4]';
detections = [0.2 0.2; 0.2 0.8; 0.7 0.1]';
dst = sqdist(sources, detections);

% a = [2 3] which means s1-d2, s2-d3
a = optimalMatchingHungarian(dst)
a2 = optimalMatchingIntProg(dst)
assert(isequal(a(:),a2(:)))

figure(1); clf
optimalMatchingDemoPlot(sources, detections, a)



%%%% Flip roles of sources and detections

dst  = dst';
% a = [0 1 2] which means d1-0, d2-s1, d3-s2
a = optimalMatchingHungarian(dst)

a2 = optimalMatchingIntProg(dst)
assert(isequal(a(:),a2(:)))

figure(2); clf
optimalMatchingDemoPlot(detections, sources, a)




%%%%%%%%%% Move s1 nearer to d1
%  d2  
%         s2 d3
%  s1 d1

sources = [0.1 0.3; 0.6 0.4]';
detections = [0.2 0.2; 0.2 0.8; 0.7 0.1]';
dst = sqdist(sources, detections);

% a = [2 3] which means s1-d2, s2-d3
a = optimalMatchingHungarian(dst)
[a2, ass] = optimalMatchingIntProg(dst)
assert(isequal(a(:),a2(:)))


figure(3); clf
optimalMatchingDemoPlot(sources, detections, a)