how to find out the exact threshold value for the correct recognition of person

Question

hello I am novice here in matlab my task is to write a code for multimodal biometric system using discrete wavelet transform and pca (principle component analysis) for 100 people database containing 9 face, 2 ear and 2 thumbprint images. for same I wrote following code for each sample

E11=imread('E1(1).JPG');
E11=imresize(E11,[100 100]);
E11=rgb2gray(E11);
%colormap(pink)
[a,b,c,d]=dwt2(E11,'haar');
X=a;E11=X(:);
E11=im2double(E11);
for reading all those images

after that i applied pca on all these images and after it i used the minimum distance classifier mindist=dist(i);

for i=2:10
    if dist(i)< mindist
        mindist=dist(i);
    end
end

for i=1:10
    if dist(i)== mindist
        img=i;
        break;
    end   
end

disp(['Recognized image number as class................']);
disp(img);

if dist(img)>=600;
    img=i;
end

like this of course before this i had found weight vectors and all after this i had wrote switch case for identifying the person

figure
subplot(1,2,1)
imshow(y), title('Image to be Recognized'),colormap(gray);
subplot(1,2,2)
x1=imresize(x1,[100,100]);
imshow(x1), title('Recognized Image'),colormap(gray);
toc
and the above last code is for recognising that image 

now my question is al this code works well for 5 classes showing correct output for 70%(no.of correct images/total no. of images) but i have to extend this code for 100 people it shows very less accuracy how can i do this please guide me i will be very grateful to you for same