diff --git a/balanceClasses.m b/balanceClasses.m
index 3ccbd98..c5574e4 100644
--- a/balanceClasses.m
+++ b/balanceClasses.m
@@ -5,7 +5,7 @@
         r=floor(size(mostFrequent,2)*rand(1)+1);
         delIndex=mostFrequent(r);
         trainClasses=trainClasses([1:delIndex-1,delIndex+1:end]);
-        trainData=trainData([1:delIndex-1,delIndex+1:end]);
+        trainData=trainData([1:delIndex-1,delIndex+1:end],:);
     end
     assert(size(trainClasses,2)>0)
 end
\ No newline at end of file
diff --git a/classifyAccordingToEMG.m b/classifyAccordingToEMG.m
new file mode 100644
index 0000000..79cdd2e
--- /dev/null
+++ b/classifyAccordingToEMG.m
@@ -0,0 +1,15 @@
+function [classification]=classifyAccordingToEMG(trainingDataEEG, trainingDataEMG, classes,shift,frequency,threshold)
+    classification=zeros([size(trainingDataEEG,1),1]);
+    oldclass=0;
+    for i=1:size(trainingDataEEG,1)
+        if mean(trainingDataEMG(i,:)) > threshold % is movement
+            class=mode(classes(max(round(shift*frequency*i-frequency),1):min(round(i*shift*frequency-1),end)));
+            if class <= 5 % task is to move
+                classification(i)=class;
+                oldclass=class;
+            else % task is to rest but movement not finished
+                classification(i)=oldclass;
+            end
+        end
+    end
+end
\ No newline at end of file
diff --git a/generateTrainingData.m b/generateTrainingData.m
index 33ea013..bf0e6e2 100644
--- a/generateTrainingData.m
+++ b/generateTrainingData.m
@@ -1,27 +1,21 @@
-function [trainingDataEEG,trainingDataEMG] = generateTrainingData(splitSignals,NFFT,window,shift,params,secs)
+function [trainingDataEEG,trainingDataEMG] = generateTrainingData(signal,NFFT,window,shift,params)
     frequency=params.SamplingRate.NumericValue;
-    tempEEG2=zeros([32,size(splitSignals,3),NFFT/2+1]);
-    tempEMG2=zeros(fix([size(splitSignals,2)-3-32,size(splitSignals,3),fix(floor(size(splitSignals,1)/(frequency*window))-1)*window/shift+1]));
+    signalWindow=frequency*window;
+    shiftProp=window/shift;
+    tempEEG=zeros([32,fix(floor(size(signal,1)/signalWindow)-1)*shiftProp+1,NFFT/2+1]);
+    tempEMG=zeros(fix([size(signal,2)-3-32,fix(floor(size(signal,1)/signalWindow)-1)*shiftProp+1]));
     
     %Filter around 50Hz and below 2 Hz
     [A,B]= butter(2,[48 52]/(frequency/2),'stop');
     [C,D]= butter(2,2/(frequency/2),'high');
 
     parfor i=1:32 %filter single channel, w/o HEOG, Synchro and 0s
-        tempEEG=zeros(size(splitSignals,3),NFFT/2+1);
-        for j=1:size(splitSignals,3) %filter single trial
-            tempEEG(j,:)=pwelch(filter(C,D,filter(A,B,splitSignals(1:secs*frequency,i,j))),[],[],NFFT,frequency);
-        end
-        tempEEG2(i,:,:)=tempEEG;
+        tempEEG(i,:,:)=shiftingPwelch(filter(C,D,filter(A,B,signal(:,i))),frequency,window,shift,NFFT);
     end
-    trainingDataEEG=permute(tempEEG2,[2 1 3]);
+    trainingDataEEG=permute(tempEEG,[2 1 3]);
     
-    parfor i=33:size(splitSignals,2)-3
-        tempEMG=zeros(fix([size(splitSignals,3),(floor(size(splitSignals,1)/(frequency*window))-1)*window/shift+1]));
-        for j=1:size(splitSignals,3) %filter single trial
-            tempEMG(j,:)=waveformLength(filter(C,D,filter(A,B,splitSignals(:,i,j))),frequency,window,shift);
-        end
-        tempEMG2(i,:,:)=tempEMG;
+    parfor i=33:size(signal,2)-3
+        tempEMG(i-32,:,:)=waveformLength(filter(C,D,filter(A,B,signal(:,i))),frequency,window,shift);
     end
-    trainingDataEMG=permute(tempEMG2,[2 1 3]);
+    trainingDataEMG=permute(tempEMG,[2 1 3]);
 end
\ No newline at end of file
diff --git a/kfoldCV.m b/kfoldCV.m
index f49d0e6..8a4587a 100644
--- a/kfoldCV.m
+++ b/kfoldCV.m
@@ -1,5 +1,5 @@
 function [model] = kfoldCV(classification,trainingData,k,cExpMax)
-    noClasses=size(unique(classification),2);
+    noClasses=size(unique(classification),1);
     cvAccurancy=zeros(2*cExpMax+1,1);
     parfor cExp=1:2*cExpMax+1
         c=2^(cExp-cExpMax-1);
@@ -21,7 +21,7 @@
     end
     [~,maxC]=max(cvAccurancy);
     bestC=2^(maxC-cExpMax-1);
-    [balanceLeftClasses, balanceLeftOut]=balanceClasses(classification,trainingData,noClasses,0.00001);
-    model=svmtrain(balanceLeftClasses, balanceLeftOut(:,:),sprintf('-t 0 -c %f -q',bestC));
+    [balancedClasses, balanceData]=balanceClasses(classification,trainingData,noClasses,0.00001);
+    model=svmtrain(balancedClasses, balanceData(:,:),sprintf('-t 0 -c %f -q',bestC));
 
 end
\ No newline at end of file
diff --git a/plotStairs.m b/plotStairs.m
new file mode 100644
index 0000000..ee7577a
--- /dev/null
+++ b/plotStairs.m
@@ -0,0 +1,10 @@
+function [ ] = plotStairs( inVec )
+%plotStairs plots Stairs for Vector
+%   plots sum up to Index i as value for i
+    stair=zeros(size(inVec));
+    for i=1:size(classification,1)
+        stair(i)=sum(inVec(1:i));
+    end
+    plot(stairs)
+end
+
diff --git a/read.m b/read.m
index d13a0ff..4f291bc 100644
--- a/read.m
+++ b/read.m
@@ -1,7 +1,30 @@
-signal=[];
-stimulusCodes=[];
-for i=1:5
+%params
+NFFT=2048;
+window=0.2;
+shift=0.2;
+k=10;
+maxExpC=7;
+maxFile=5;
+threshold=7500;
+
+trainingDataEEGcell=cell(maxFile,1);
+trainingDataEMGcell=cell(maxFile,1);
+classesCell=cell(maxFile,1);
+
+for i=1:maxFile
     [sig, stat, params] = load_bcidat(sprintf('/home/jph/Uni/masterarbeit/Block1_ReachingMovements/AO/AO_B1001/AO_B1S001R0%i.dat',i));
-    signal=cat(1,signal,sig);
-    stimulusCodes=cat(1,stat.StimulusCode,stimulusCodes);
+    [trainingDataEEGcell{i},trainingDataEMGcell{i}]=generateTrainingData(sig,NFFT,window,shift,params);
+    classesCell{i}=stat.StimulusCode;
+    fprintf('%ith file processed\n',i)
 end
+
+trainingDataEEG=cell2mat(trainingDataEEGcell);
+trainingDataEMG=cell2mat(trainingDataEMGcell);
+classes=cell2mat(classesCell);
+
+classification=classifyAccordingToEMG(trainingDataEEG, trainingDataEMG,classes,shift,params.SamplingRate.NumericValue,threshold);
+
+smoothClassification=zeros(size(classification));
+for i=1:size(classification,1)
+    smoothClassification(i)=round(mean(classification(max(i-round(1/window),1):min(i+round(1/window),end))));        
+end
\ No newline at end of file
diff --git a/readme.m b/readme.m
index 28b94dd..9bb3767 100644
--- a/readme.m
+++ b/readme.m
@@ -53,12 +53,26 @@
 
 % Aufteilung EMG xor EEG (done)
 % danach dann Bewegungen aufteilenC
+% EMG ganze Bewegung
 
 
 % von einer Person und einem Tag zusammen (done)
 % nur die ersten zwei Sekunden der Bewegung (done [generateTrainingData])
 % für WL keine Fenster sondern ganze Bewegung (done s.o.)
-% äußere Crossvalidation
-% Confusion
+% äußere Crossvalidation (done)
+% Confusion (done)
 % Ruhe / Bewegung aufteilen EEG EMG
 %  -> testen ob mit EMG besser
+% Bewegungen mit EEG: 30.59%
+% Bewegungen mit EMG: 26.4112%
+% Move/Rest EMG: 53.7683%
+% Move/Rest EEG: 53.2012%
+
+
+% libsvm vorne in den path (done)
+% besser erst fft bzw. WL, dann aneinanderhängen (done, MEMORY!))
+% schneiden anhand von EMG (mean oder max)
+% zum Trainieren Pause vor Bewegung(= EMG-Aktivität)
+% threshold anhand von historgam/plot (bimodal?)
+
+% für confusion: caxis, normierung für predicted
diff --git a/shiftingPwelch.m b/shiftingPwelch.m
new file mode 100644
index 0000000..c0dc716
--- /dev/null
+++ b/shiftingPwelch.m
@@ -0,0 +1,9 @@
+function [EEG] = shiftingPwelch(signal,frequency,window,windowShift,NFFT)
+    signalShift=frequency*windowShift;
+    signalWindow=frequency*window;
+    shiftProp=window/windowShift;
+    EEG=zeros([fix(floor(size(signal,1)/signalWindow)-1)*shiftProp+1,NFFT/2+1]);
+    for j=1:fix(floor(size(signal,1)/signalWindow-1)*shiftProp+1)
+        EEG(j,:)=pwelch(signal((j-1)*signalShift+1:(j-1)*signalShift+signalWindow),[],[],NFFT,2500);
+    end
+end
\ No newline at end of file
diff --git a/svm.m b/svm.m
index 62cc6cf..f298ba0 100644
--- a/svm.m
+++ b/svm.m
@@ -1,53 +1,41 @@
+close all
+clear all
+clc
+
 %params
 NFFT=2048;
 cutWindowEEG=2;
 window=0.2;
 shift=0.05;
-k=10;
-rounds=1;
-maxExpC=7;
 
-%svm
 load('/home/jph/Uni/masterarbeit/data/AO1.mat');
 
-[ splitSignals, code ] = cutMovements( stimulusCodes, signal );
-%[ splitSignals, code ] = cutWindows( stimulusCodes, signal, params, cutWindow );
-
-%trainingData has format trial x ch x powerSpectrum
-[trainingDataEEG,trainingDataEMG]=generateTrainingData(splitSignals,NFFT,window,shift,params,cutWindowEEG);
-
-%classify for different movements and rest
-classification=zeros(size(code));
-for i=1:size(code)
-    if code(i)<=5
-        classification(i)=code(i);
-    else
-        classification(i)=0;
-    end
-end
-
-
-%classify for move/rest
-% cann't use cutMovements then
-%classification=double(code<=5);
+%k=2; 
+maxExpC=5; % c\in {2^i|i=-maxExpC:1:maxExpC}
 
 %choose to estimate based on EEG or EMG
-trainingData=trainingDataEEG;
+trainingData=trainingDataEMG;
+clear trainingDataEEG;
+clear trainingDataEMG;
 
-
-accurancy=zeros(rounds,3);
-means=zeros(rounds,1);
+accurancy=zeros(k,3);
+noClasses=size(unique(classification),1);
+cm=zeros(noClasses);
+randMap=randperm(size(trainingData,1));
+disp('startCV')
 for i=1:k
-    randMap=randperm(size(trainingData,1));
-    leaveOut=trainingData(mod(randMap,k+1)==i,:,:);
-    leaveClasses=classification(mod(randMap,k+1)==i);
-    remaining=trainingData(mod(randMap,k+1)~=i,:,:);
-    remainingClasses=classification(mod(randMap,k+1)~=i);
-
+    leaveOut=trainingData(mod(randMap,k)==i-1,:,:);
+    leaveClasses=classification(mod(randMap,k)==i-1);
+    remaining=trainingData(mod(randMap,k)~=i-1,:,:);
+    remainingClasses=classification(mod(randMap,k)~=i-1);
+    disp(datestr(datetime('now')))
+    fprintf('create %ith model\n',i)
     model=kfoldCV(remainingClasses,remaining,k,maxExpC);
 
+    disp(datestr(datetime('now')))
     [predictions,accurancy(i,:),pvalues]=svmpredict(leaveClasses,leaveOut(:,:),model);
-    means(i)=mean(leaveClasses-predictions); %confusion matrix
+    cm=cm+confusionmat(leaveClasses,predictions); %confusion matrix
     %cf: for each cue what was the prediction (maybe colourCoded)
+    %fprintf('%ith step done\n',i)
 end
 meanAccurancy=mean(accurancy(:,1))
\ No newline at end of file
diff --git a/text/collection.txt b/text/collection.txt
new file mode 100644
index 0000000..7025eb6
--- /dev/null
+++ b/text/collection.txt
@@ -0,0 +1,14 @@
+Results
+=======
+given Classification
+--------------------
+Bewegungen mit EEG: 30.59%
+Bewegungen mit EMG: 26.4112%
+Move/Rest EMG: 53.7683%
+Move/Rest EEG: 53.2012%
+
+new Classification
+------------------
+Move/Rest EMG : 99.9535%
+Bewegungen EMG: 74.4832% *k=2, c=1*
+
diff --git a/waveformLength.m b/waveformLength.m
index 8239c05..3b927f3 100644
--- a/waveformLength.m
+++ b/waveformLength.m
@@ -5,8 +5,8 @@
     signalWindow=frequency*window;
     shiftProp=window/windowShift;
     WL=zeros((floor(size(signal,1)/signalWindow)-1)*shiftProp+1,1);
-    for j=1:(size(signal,1)/signalWindow-1)*shiftProp+1
-        WL(j)=sum(diff(signal((j-1)*signalShift+1:(j-1)*signalShift+signalWindow)));
+    for j=1:fix(floor(size(signal,1)/signalWindow-1)*shiftProp+1)
+        WL(j)=sum(abs(diff(signal((j-1)*signalShift+1:(j-1)*signalShift+signalWindow))));
     end
 end