weka文件生成脚

从findsound中选取了8个大类，并且对每个大类中的5个小类（一个为三个小类）进行了音频特征的提取，至多提取100个clip。提取的特征以mean和variance的形式分类保存在了文件之中，并且每一个clip对应有相应的文件，这些文件和相应的clip保存在相应的目录之中。

功能：根据给定的特征的类型，从这些文件中选择相应特征值，组成针对每一个小类的arff weka文件，这些weka文件保存到一个统一的地方，这里是保存在了list所在的目录。

list文件也是针对每一个小类的，并且按照原有目录的层次结构进行了组织。

function [ret,contents] = parseAfile(fpath,target)state = exist(fpath,'file');if state~=2   sprintf('not exist %s \n',fpath)   ret=0;   contents=cell(1,1);   return;endcontents=cell(1,length(target));fid=fopen(fpath);count=1;while ~feof(fid)  featurename=fgetl(fid);  featurename=strtrim(featurename);  l=length(target);  flag=0;  for i=1:l     name=target(i);     if strcmp(name,featurename)         flag=1;         break;     end  end  if flag==0      continue;  end  featurevalue=fgetl(fid);  featurevalue=strtrim(featurevalue);  featurevalue=sprintf('%s ',featurevalue);  index=strfind(featurevalue,' ');  aline=cell(1,length(index));  start=1;  for j=1:length(index)     avalue=featurevalue(1,start:index(j));     start=index(j)+1;     aline{j}=avalue;   end  contents{count}=aline;

  count=count+1;endfclose(fid);ret=1;

上面的代码根据fpath指定的路径，找到特征文件，并且根据target返回所需要的特征值。

alllist=cell(8,5);alllist{1,1}='animal\bat.txt';alllist{1,2}='animal\cat.txt';alllist{1,3}='animal\cow.txt';alllist{1,4}='animal\dog.txt';alllist{1,5}='animal\lamb.txt';alllist{2,1}='household\bubbles.txt';alllist{2,2}='household\clock.txt';alllist{2,3}='household\door.txt';alllist{2,4}='household\phone.txt';alllist{2,5}='household\toilet.txt';alllist{3,1}='musical+instruments\drum+rimshot.txt';alllist{3,2}='musical+instruments\flute.txt';alllist{3,3}='musical+instruments\guitar.txt';alllist{3,4}='musical+instruments\piano.txt';alllist{3,5}='musical+instruments\trumpet.txt';alllist{4,1}='nature\fire.txt';alllist{4,2}='nature\ocean.txt';alllist{4,3}='nature\rain.txt';alllist{4,4}='nature\thunder.txt';alllist{4,5}='nature\wind.txt';alllist{5,1}='office\coins.txt';alllist{5,2}='office\modem.txt';alllist{5,3}='office\mouse+click.txt';alllist{5,4}='office\paper.txt';alllist{5,5}='office\typewriter.txt';alllist{6,1}='people\applause.txt';alllist{6,2}='people\baby.txt';alllist{6,3}='people\cough.txt';alllist{6,4}='people\cry.txt';alllist{6,5}='people\heartbeat.txt';alllist{7,1}='sports+and+recreation\bowling.txt';alllist{7,2}='sports+and+recreation\camera.txt';alllist{7,3}='sports+and+recreation\cards.txt';alllist{8,1}='vehicles\collision.txt';alllist{8,2}='vehicles\engine.txt';alllist{8,3}='vehicles\helicopter.txt';alllist{8,4}='vehicles\ship.txt';alllist{8,5}='vehicles\train.txt';[r,c]=size(alllist);target=cell(1,10);target{1,1}='fluctuation';target{1,2}='attacktime';target{1,3}='attackslope';target{1,4}='centroid';target{1,5}='brightness';target{1,6}='skewness';target{1,7}='kurtosis';target{1,8}='spectentropy';target{1,9}='flatness';target{1,10}='irregularity';target{1,11}='zerocross';target{1,12}='mfcc';allcategory=cell(1,38);allcategory{1,1}='bat';allcategory{1,2}='cat';allcategory{1,3}='cow';allcategory{1,4}='dog';allcategory{1,5}='lamb';allcategory{1,6}='bubbles';allcategory{1,7}='clock';allcategory{1,8}='door';allcategory{1,9}='phone';allcategory{1,10}='toilet';allcategory{1,11}='drum+rimshot';allcategory{1,12}='flute';allcategory{1,13}='guitar';allcategory{1,14}='piano';allcategory{1,15}='trumpet';allcategory{1,16}='fire';allcategory{1,17}='ocean';allcategory{1,18}='rain';allcategory{1,19}='thunder';allcategory{1,20}='wind';allcategory{1,21}='coins';allcategory{1,22}='modem';allcategory{1,23}='mouse+click';allcategory{1,24}='paper';allcategory{1,25}='typewriter';allcategory{1,26}='applause';allcategory{1,27}='baby';allcategory{1,28}='cough';allcategory{1,29}='cry';allcategory{1,30}='heartbeat';allcategory{1,31}='bowling';allcategory{1,32}='camera';allcategory{1,33}='cards';allcategory{1,34}='collision';allcategory{1,35}='engine';allcategory{1,36}='helicopter';allcategory{1,37}='ship';allcategory{1,38}='train';hasattribute=0;for i=1:r    for j=1:c        alist=alllist{i,j}        if isempty(alist)           continue;         end        aalist=sprintf('list2\\%s',alist)        fid = fopen(aalist);        output=sprintf('list2\\%s.arff',alist);        fidoutput=fopen(output,'w+')        fprintf(fidoutput,'@relation ''cpu''\n');        while ~feof(fid)           line=fgetl(fid);           line=sprintf('%s-mean.txt',line);           indexs=strfind(line,'/');           len=length(indexs);           category=line(1,indexs(len-1)+1:indexs(len)-1)           fullpath=sprintf('F:\\研究——音频事件检测\\findsound  result\\%s',line)                     [ret,contents]=parseAfile(fullpath,target);           if ret==1               if hasattribute==0                   [retstat,order]=getattributename(fullpath,target);                   for k=1:length(order)                      name=order{k,1};                      le=order{k,2};                      for kk=1:le                         %thisname=sprintf('%s%d\n',name,kk)                         %fwrite(fidoutput,thisname);                         fprintf(fidoutput,'@attribute %s%d numeric\n',name,kk);                      end                   end                   fprintf(fidoutput,'@attribute class {')                   for kk=1:length(allcategory)                       if kk~=length(allcategory)                           fprintf(fidoutput,'%s,',allcategory{kk});                           continue;                       end                       fprintf(fidoutput,'%s',allcategory{kk});                   end                   fprintf(fidoutput,'}\n');                   fprintf(fidoutput,'@data\n');                   hasattribute=1;               end                              tmp=contents;               for k=1:length(contents)                  avalue=contents{1,k};                  for kk=1:length(avalue)                     %fwrite(fidoutput,avalue{kk});                     %fwrite(fidoutput,' ');                     fprintf(fidoutput,'%s,',avalue{kk});                  end               end               %fwrite(fidoutput,'\n');               fprintf(fidoutput,'%s\n',category);           end        end        hasattribute=0;        fclose(fid);        fclose(fidoutput);    endend

上面的代码，指定了所有的list的位置，所有的类别，所要提取的特征的类型，通过调用脚本一，得到这些特征，并且把这些特征输出为arff文件

function [ret,orderattribute] = getattributename(fpath,target)state = exist(fpath,'file');if state~=2   sprintf('not exist %s \n',fpath)   ret=0;   orderattribute=cell(1,1);   return;endorderattribute=cell(length(target),2);fid=fopen(fpath);count=1;while ~feof(fid)  featurename=fgetl(fid);  featurename=strtrim(featurename);  l=length(target);  flag=0;  for i=1:l     name=target(i);     if strcmp(name,featurename)         flag=1;         break;     end  end  if flag==0      continue;  end  featurevalue=fgetl(fid);  featurevalue=strtrim(featurevalue);  featurevalue=sprintf('%s ',featurevalue);  index=strfind(featurevalue,' ');  attrlen=length(index);  orderattribute{count,1}=featurename;  orderattribute{count,2}=attrlen;  count=count+1;endfclose(fid);ret=1;

getattributename的功能是返回所有的特征，并且按照实际的顺序，不应是按照target指定的顺序，可以作为arff文件中的attribute的输出