%?2018年11月22日?by?zhangyaokun
%?BP?identification
clear?all;
close?all;
%%?語音數(shù)據(jù)預(yù)處理方法:先隨機(jī)打亂數(shù)據(jù)，再對(duì)數(shù)據(jù)進(jìn)行歸一化，然后可以直接順序選取訓(xùn)練數(shù)據(jù)和測(cè)試數(shù)據(jù)
%?load?data1.mat;?load?data2.mat;?load?data3.mat;?load?data4.mat;?%加載原數(shù)據(jù)
%?Ori_Data1=[c1;c2;c3;c4];??%全部的原始數(shù)據(jù)?2000*25
%?rowrank?=?randperm（size（Ori_Data1?1））;?%?隨機(jī)打亂的數(shù)字，從1~行數(shù)打亂
%?Ori_Data?=?Ori_Data1（rowrank?:）;%%按照rowrank打亂矩陣的行數(shù)?2000*25
%?
%?%輸入數(shù)據(jù)處理：歸一化
%?Ori_Data_X?=?Ori_Data（:2:25）;??%取出原始數(shù)據(jù)中的特征向量?2000*24
%?FlattenedData?=?Ori_Data_X（:）‘;?%?展開矩陣為一列，然后轉(zhuǎn)置為一行。
%?MappedFlattened?=?mapminmax（FlattenedData）;?%?對(duì)整個(gè)矩陣的數(shù)據(jù)進(jìn)行歸一化，到[-11]之間
%?MappedData?=?reshape（MappedFlattened?size（Ori_Data_X））;?%?還原為原始矩陣形式。此處不需轉(zhuǎn)置回去，因?yàn)閞eshape恰好是按列重新排序??2000*24
%?X=MappedData（1:1500:）‘;?Xc=MappedData（1501:2000:）‘;?%前1500個(gè)數(shù)據(jù)全部作為訓(xùn)練數(shù)據(jù)，后500個(gè)數(shù)據(jù)作為測(cè)試數(shù)據(jù)
%?
%?%輸出數(shù)據(jù)處理：%定義輸出向量?分別為[1?0?0?0]?[0?1?0?0]?[0?0?1?0]?[0?0?0?1]
%?To=?Ori_Data（:1）‘;%取出第一列元素為輸出
%?Tt=[1?0?0?0];??????????????????????????????????????????????????????????
%?for?i=1:1:2000
%?????switch?To（i）
%?????????case?1
%?????????????Tt（i:）=[1?0?0?0];
%?????????case?2
%?????????????Tt（i:）=[0?1?0?0];??
%?????????case?3
%?????????????Tt（i:）=[0?0?1?0];
%?????????case?4
%?????????????Tt（i:）=[0?0?0?1];????
%?????end
%?end
%?T=Tt（1:1500:）‘;???Tc=Tt（1501:2000:）‘;
%?save（‘traindata_X‘‘X‘）;?save（‘traindata_Xc‘‘Xc‘）;?%保存訓(xùn)練數(shù)據(jù)
%?save（‘traindata_To‘‘To‘）;save（‘traindata_T‘‘T‘）;?save（‘traindata_Tc‘‘Tc‘）;??%保存測(cè)試數(shù)據(jù)??
%?
%?clearvars?-except?X?T?Xc?Tc?To;??%清除其余變量，僅保留訓(xùn)練數(shù)據(jù)和測(cè)試數(shù)據(jù)
%上面的程序只需要運(yùn)行一遍
?load?traindata_X;?load?traindata_T;?load?traindata_Xc;?load?traindata_Tc;?load?traindata_To;%加載訓(xùn)練數(shù)據(jù)和測(cè)試數(shù)據(jù)
?%%?BP網(wǎng)絡(luò)結(jié)構(gòu)及參數(shù)初始化
?%?BP網(wǎng)絡(luò)結(jié)構(gòu)為?24-10-4
?Hid_num?=?10;
?Func=0;
?
[RQ]=?size（X）;
[S~]=?size（T）;
[ScQc]=?size（Tc）;
%%?使用不同的隱含層數(shù)目訓(xùn)練網(wǎng)絡(luò)??
%?for?k=5:5:40?%不同的隱含層數(shù)目
%?????Hid_num?=?k;
%?????
%?????net=newff（XTHid_num）;??%建立BP神經(jīng)網(wǎng)絡(luò)?{‘logsig‘‘purelin‘}‘trainlm‘
%?????switch?Func??%選擇訓(xùn)練函數(shù)
%?????????case?1
%??????????????net.?trainFcn=‘traingd‘;
%?????????case?2
%??????????????net.?trainFcn=‘traingdm‘;
%?????????case?3
%??????????????net.?trainFcn=‘traingdx‘;
%?????????otherwise
%??????????????net.?trainFcn=‘trainlm‘;
%?????end?
%?
%?????net.trainParam.epochs=2000;??%設(shè)置訓(xùn)練次數(shù)
%?????net.trainParam.max_fail?=?50;
%?????net.trainParam.goal=0.005;??%設(shè)置mean?square?error，?均方誤差
%?????net.trainParam.lr=0.03;?%設(shè)置學(xué)習(xí)速率
%?
%?????net.iw{11}=rands（Hid_numR）;??%隱含層神經(jīng)元的初始權(quán)值
%?????net.lw{21}=rands（SHid_num）;??%輸出層神經(jīng)元的初始權(quán)值
%?????net.b{1}=rands（Hid_num1）;??%隱含層神經(jīng)元的初始偏置
%?????net.b{2}=rands（S1）;??%輸出層神經(jīng)元的初始偏置
%?????[nettr_gd]=train（netXT）;??%訓(xùn)練網(wǎng)絡(luò)
%?
%?????A=sim（netX）;??%測(cè)試網(wǎng)絡(luò)
%?????E=T-A;??%測(cè)試誤差
%?????error?=?sumsqr（E）/（S*Q）??%測(cè)試結(jié)果的的MSE
%?????A1=sim（netXc）;??%測(cè)試網(wǎng)絡(luò)
%?????E1=Tc-A1;??%測(cè)試誤差
%?????error_c1?=?sumsqr（E1）/（Sc*Qc）??%測(cè)試結(jié)果的的MSE
%?????for?i=1:500??????
%?????????Res（i）=find（A1（:i）==max（A1（:i）））;?%以能夠識(shí)別出最大值的趨勢(shì)，作為正確識(shí)別的標(biāo)準(zhǔn)
%?????end
%?????E2=To（:1501:2000）-Res;
%?????I=find（E2（:）~=0）;??%識(shí)別錯(cuò)誤的數(shù)據(jù)
%?????length（I）?????%識(shí)別錯(cuò)誤的個(gè)數(shù)
%?????
%?????pause
%?en

?屬性????????????大小?????日期????時(shí)間???名稱
-----------?---------??----------?-----??----
?????目錄???????????0??2019-04-25?02:29??BP_Music_recog\
?????文件???????22690??2018-11-28?12:08??BP_Music_recog\20_elm.JPG
?????文件???????20147??2018-11-28?12:10??BP_Music_recog\20_gd.JPG
?????文件???????20128??2018-11-28?12:11??BP_Music_recog\20_gdm.JPG
?????文件???????19530??2018-11-28?12:12??BP_Music_recog\20_gdx.JPG
?????文件????????4994??2018-11-28?12:33??BP_Music_recog\BP_Music_recog.m
?????文件???????93015??2009-08-29?16:54??BP_Music_recog\data1.mat
?????文件???????92845??2009-08-29?16:54??BP_Music_recog\data2.mat
?????文件???????92937??2009-08-29?16:54??BP_Music_recog\data3.mat
?????文件???????93438??2009-08-29?16:54??BP_Music_recog\data4.mat
?????文件???????22569??2018-11-28?11:48??BP_Music_recog\elm_10.JPG
?????文件???????22082??2018-11-28?11:49??BP_Music_recog\elm_15.JPG
?????文件???????22581??2018-11-28?11:51??BP_Music_recog\elm_20.JPG
?????文件???????22720??2018-11-28?11:52??BP_Music_recog\elm_25.JPG
?????文件???????22978??2018-11-28?12:01??BP_Music_recog\elm_30.JPG
?????文件???????22507??2018-11-28?11:55??BP_Music_recog\elm_35.JPG
?????文件???????22085??2018-11-27?16:59??BP_Music_recog\elm_40.JPG
?????文件???????22867??2018-11-28?11:47??BP_Music_recog\elm_5.JPG
?????文件???????25454??2018-11-27?17:02??BP_Music_recog\error.jpg
?????文件?????????912??2018-11-28?11:31??BP_Music_recog\traindata_T.mat
?????文件?????????435??2018-11-28?11:31??BP_Music_recog\traindata_Tc.mat
?????文件?????????854??2018-11-28?11:31??BP_Music_recog\traindata_To.mat
?????文件??????273878??2018-11-28?11:31??BP_Music_recog\traindata_X.mat
?????文件???????91390??2018-11-28?11:31??BP_Music_recog\traindata_Xc.mat
?????文件????????2458??2018-11-28?10:10??BP_Music_recog\work3.m
?????文件??????191558??2019-04-25?02:28??語音特征識(shí)別-神經(jīng)網(wǎng)絡(luò)應(yīng)用.docx