clc

%?第一步，利用主元分析PCA法提取圖像的特征矩陣
allsamples=[];????????????????????????????????%所有訓練圖像
for?i=1:15
????for?j=1:5
??????a=imread（strcat（pwd‘\Yale\s‘num2str（i）‘\‘num2str（j）‘.bmp‘））;
??????b=a（1:100*80）;??????????????????????????
??????b=double（b）;
??????allsamples=[allsamples;?b];????????????
????end
end
samplemean=mean（allsamples）;??????????????????
for?i=1:75
????xmean（i:）=allsamples（i:）-samplemean;????
end
sigma=xmean*xmean‘;???????
[v?d]=eig（sigma）;
d1=diag（d）;
[d2?index1]=sort（d1）;?????
cols=size（v2）;???????????
for?i=1:cols
????vsort（:i）?=?v（:?index1（cols-i+1）?）;???
????dsort（i）???=?d1（?index1（cols-i+1）?）;????
end????%完成降序排列
%以下選擇85%的能量
dsum?=?sum（dsort）;
????dsum_extract?=?0;
????p=?0;
????while（?dsum_extract/dsum?????????p=p?+?1;
????????dsum_extract?=?sum（dsort（1:p））;
????end
i=1;
%計算特征臉形成的坐標系
while?（i<=p?&&?dsort（i）>0）
????base（:i）?=?dsort（i）^（-1/2）?*?xmean‘?*?vsort（:i）;?????%?base是N×p階矩陣，除以dsort（i）^（1/2）是對人臉圖像的標準化
????i?=?i?+?1;
end

%下面的代碼，將訓練樣本對坐標系上進行投影得到一個?M*p?階矩陣allcoor
allcoor?=?allsamples?*?base;

%第二步，創(chuàng)建并訓練BP神經(jīng)網(wǎng)絡
%生成訓練BP神經(jīng)網(wǎng)絡的輸入?P?
P=?mapminmax（allcoor）;???????????????%歸一化

%生成目標輸出矢量?T
T=zeros（7515）;
?for?i=1:15
????for?j=1:5
??????T（（i-1）*5+ji）=1;
??end
?end
?
%打亂訓練樣本順序
gx2（:1:19）=P;
gx2（:20:34）=T;
xd=gx2（randperm（numel（gx2）/34）:）;
gx=xd（:1:19）;d=xd（:20:34）;
P=gx‘;
T=d‘;

%創(chuàng)建BP神經(jīng)網(wǎng)絡
[RQ]=size（P）;
[S2Q]=size（T）;
net=newcf（minmax（P）T[40]{‘logsig‘}‘trainscg‘）;

%訓練BP神經(jīng)網(wǎng)絡
net.trainparam.epochs=5000;
net.trainparam.goal=0.0001;
net.divideFcn?=?‘‘;
net=train（netPT）;

%仿真BP神經(jīng)網(wǎng)絡
Y=sim（netP）;

%?第三步，測試BP神經(jīng)網(wǎng)絡并計算其識別率
%測試BP神經(jīng)網(wǎng)絡
s=0;
for?i=1:15
????for?j=6:10????????????????????????????????%讀入15x5副測試圖像
?????????a=imread（strcat（pwd‘\Yale\s‘num2str（i）‘\‘num2str（j）‘.bmp‘））;
?????????b=a（1:8000）;
?????????b=double（b）;
?????????tcoor=?b?*?base;???????????????????%計算坐標，是1×p階矩陣
?????????X?=?mapminmax（tcoor）;??????????????%歸一化
?????????Z=sim（netX‘）;
?????????[ziindex2]=max（Z）;
?????????if?index2==i???
?????????????s=s+1;

??????????end
?????end
end

%計算識別率
accuracy=s/Q;
fprintf（2‘采用PCA和神經(jīng)網(wǎng)絡的識別率為：%.3f%%\n\n‘a(chǎn)ccuracy*100）;

?屬性????????????大小?????日期????時間???名稱
-----------?---------??----------?-----??----

?????文件???????2410??2017-12-28?13:18??PCA+BP\pcabp.m

?????文件????????271??2017-12-28?20:19??PCA+BP\readme.txt

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s1\1.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s1\10.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s1\11.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s1\2.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s1\3.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s1\4.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s1\5.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s1\6.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s1\7.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s1\8.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s1\9.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s10\1.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s10\10.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s10\11.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s10\2.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s10\3.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s10\4.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s10\5.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s10\6.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s10\7.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s10\8.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s10\9.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s11\1.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s11\10.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s11\11.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s11\2.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s11\3.bmp

?????文件???????9080??2017-11-01?23:06??PCA+BP\Yale\s11\4.bmp

............此處省略159個文件信息