diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..b25c15b
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1 @@
+*~
diff --git a/2015-sarasola-sanz-EMG-based multi-joint kinematics decoding for robot-aided rehabilitation therapies.pdf b/2015-sarasola-sanz-EMG-based multi-joint kinematics decoding for robot-aided rehabilitation therapies.pdf
new file mode 100644
index 0000000..077f0c6
--- /dev/null
+++ b/2015-sarasola-sanz-EMG-based multi-joint kinematics decoding for robot-aided rehabilitation therapies.pdf
Binary files differ
diff --git a/2015-shiman-Towards Decoding of Functional Movements from the Same Limb using EEG.pdf b/2015-shiman-Towards Decoding of Functional Movements from the Same Limb using EEG.pdf
new file mode 100644
index 0000000..ac26fcf
--- /dev/null
+++ b/2015-shiman-Towards Decoding of Functional Movements from the Same Limb using EEG.pdf
Binary files differ
diff --git a/2016-spueler-Comparing Methods for Decoding Movement Trajectory from ECoG in Chronic Stroke Patients.pdf b/2016-spueler-Comparing Methods for Decoding Movement Trajectory from ECoG in Chronic Stroke Patients.pdf
new file mode 100644
index 0000000..33278f5
--- /dev/null
+++ b/2016-spueler-Comparing Methods for Decoding Movement Trajectory from ECoG in Chronic Stroke Patients.pdf
Binary files differ
diff --git a/analyze.m b/analyze.m
new file mode 100644
index 0000000..dbcc954
--- /dev/null
+++ b/analyze.m
@@ -0,0 +1,24 @@
+[signal, states, params] = load_bcidat('/home/jph/Uni/Masterarbeit/Block1_ReachingMovements/AO/AO_B1001/AO_B1S001R01.dat');
+
+% [ splitSignals, code ] = cutAll( states, signal );
+[ splitSignals, code ] = cutWindows( states, signal, params, 1);
+
+NFFT=2048; % should be 2^n
+
+[ filteredSig, pwMeanMove, pwMeanRest,hz ] = myFilter( splitSignals,code,NFFT );
+
+%plotChannelProzDiff(pwMeanMove,pwMeanRest,hz,params);
+%plotChannels(pwMeanMove,hz,params);
+plotImgScProzDiff( pwMeanMove, pwMeanRest)
+
+% figure
+% subplot(2,1,1)
+% plot(log(pwMeanMove))
+% xlim([0 150])
+% %ylim([-2 6])
+% subplot(2,1,2)
+% plot(log(pwMeanRest))
+% xlim([0 150])
+% %ylim([-2 6])
+
+    
\ No newline at end of file
diff --git a/balance2Classes.m b/balance2Classes.m
new file mode 100644
index 0000000..f4bae25
--- /dev/null
+++ b/balance2Classes.m
@@ -0,0 +1,14 @@
+function [trainClasses,trainData]=balance2Classes(trainClasses,trainData)
+    %balances two classes (0,1)
+    eps=0.00001;
+    while mean(trainClasses)> 0.5+eps || mean(trainClasses) < 0.5-eps
+        r=floor(0.5*size(trainClasses)*rand(1)+1);
+        if mean(trainClasses) < 0.5
+            more=find(trainClasses==0);
+        else
+            more=find(trainClasses==1);
+        end
+        trainClasses=trainClasses([1:more(r)-1,more(r)+1:end]);
+        trainData=trainData([1:more(r)-1,more(r)+1:end],:);
+    end
+end
\ No newline at end of file
diff --git a/balanceClasses.m b/balanceClasses.m
new file mode 100644
index 0000000..e389592
--- /dev/null
+++ b/balanceClasses.m
@@ -0,0 +1,10 @@
+function [trainClasses,trainData]=balanceClasses(trainClasses,trainData,eps)
+    while sum(abs(diff(histcounts(trainClasses,size(unique(trainClasses),2)))))> 0+eps
+        mostFrequent=find(trainClasses==mode(trainClasses));
+        
+        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]);
+    end
+end
\ No newline at end of file
diff --git a/chooseForCode.m b/chooseForCode.m
new file mode 100644
index 0000000..aa15422
--- /dev/null
+++ b/chooseForCode.m
@@ -0,0 +1,12 @@
+function [ splitSamplesCode ] = chooseForCode(code,codes, splitSamples)
+%chooseForCode chooses the samples for given code
+%   input the code to choose, the list of codes, the number of Samples and
+%   the list of Samples to get the list with the given code
+    splitSamplesCode=[];
+    for i=1:size(codes)
+        if code == codes(i)
+            splitSamplesCode=[splitSamplesCode,splitSamples(:,:,i)];
+        end
+    end
+end
+
diff --git a/cutAll.m b/cutAll.m
new file mode 100644
index 0000000..123d7b6
--- /dev/null
+++ b/cutAll.m
@@ -0,0 +1,16 @@
+function [ splitSignals, code ] = cutAll( states, signal )
+% cut all in same length by dropping the tail
+
+changeIndices=[find(diff(double(states.StimulusCode)));size(states.StimulusCode,1)];
+codeCount=[diff(changeIndices); size(states.StimulusCode,1)-sum(diff(changeIndices))];
+
+splitSignals=[];
+code =zeros(size(codeCount,1));
+c=codeCount(1);
+
+for i=2:size(codeCount,1)
+    splitSignals=cat(3,splitSignals,signal(c:c+4998,:));
+    code(i)=states.StimulusCode(c);
+    c=c+codeCount(i);
+end
+end
\ No newline at end of file
diff --git a/cutTrials.m b/cutTrials.m
new file mode 100644
index 0000000..18b87bd
--- /dev/null
+++ b/cutTrials.m
@@ -0,0 +1,22 @@
+function [ splitSignals, code ] = cutTrials( states, signal )
+%cut Trials only, ignore resting phases
+
+noSamplesPerTrial=15000;
+% count how many recorded values belong to one trial
+changeIndices=[find(diff(double(states.StimulusCode)));size(states.StimulusCode,1)];
+codeCount=[diff(changeIndices); size(states.StimulusCode,1)-sum(diff(changeIndices))];
+
+splitSignals=[];
+code=[];
+
+c=1;
+
+for i=1:size(codeCount,1)
+    if(codeCount(i)==noSamplesPerTrial)
+        splitSignals=cat(3,splitSignals,signal(c:c+codeCount(i)-1,:));
+        code=cat(1,code,states.StimulusCode(c));
+    end
+    c=c+codeCount(i);
+end
+end
+
diff --git a/cutWindows.m b/cutWindows.m
new file mode 100644
index 0000000..44b50d6
--- /dev/null
+++ b/cutWindows.m
@@ -0,0 +1,21 @@
+function [ splitSignals, code ] = cutWindows( states, signal, params, t )
+    changeIndices=cat(1,0,find(diff(double(states.StimulusCode))));
+    codeCount=diff(changeIndices);
+    interval=t*params.SamplingRate.NumericValue;
+
+    splitSignals=[];
+    code=[];
+
+    c=codeCount(1)+1;
+    %cut windows of 1s
+    for i=2:(size(codeCount,1)) %exclude beginning and end (stimulus 0)
+        tempc=0;
+        while codeCount(i)-tempc >= interval
+            splitSignals=cat(3,splitSignals,signal(c+tempc:c+tempc+2499,:));
+            code=cat(1,code,double(states.StimulusCode(c+tempc)));
+            tempc=tempc+interval;
+        end
+        c=c+codeCount(i);
+    end
+end
+
diff --git a/data/accurancys15.mat b/data/accurancys15.mat
new file mode 100644
index 0000000..b9bb82b
--- /dev/null
+++ b/data/accurancys15.mat
Binary files differ
diff --git a/data/accurancysInCV.mat b/data/accurancysInCV.mat
new file mode 100644
index 0000000..7884903
--- /dev/null
+++ b/data/accurancysInCV.mat
Binary files differ
diff --git a/fileInfo.txt b/fileInfo.txt
new file mode 100644
index 0000000..a4a7f43
--- /dev/null
+++ b/fileInfo.txt
@@ -0,0 +1,11 @@
+Stimuli Sequence
+12 3 8 1 10 4 7 2 15 4 12 1 14 2 9 3 11 2 13 4 10 3 7 3 12 1 9 1 15 1 6 3 13 2 8 4 11 1 14 4 6 4 9 1 14 2 7 4 6 1 8 3 10 3 7 2 15 4 12 1 14 2 9 4 11 3 13 4 10 2 7 1 12 3 9 2 15 2 6 3 13
+
+Pre & Post Run
+3s
+
+Stimulus duration
+4s
+
+Sampling Rate
+2500
diff --git a/generateTrainingData.m b/generateTrainingData.m
new file mode 100644
index 0000000..a467afc
--- /dev/null
+++ b/generateTrainingData.m
@@ -0,0 +1,27 @@
+function [trainingDataEEG,trainingDataEMG] = generateTrainingData(splitSignals,NFFT,window,shift,params)
+    frequency=params.SamplingRate.NumericValue;
+    tempEEG2=zeros([32,size(splitSignals,3),NFFT/2+1]);
+    tempEMG2=zeros([size(splitSignals,2)-3-32,size(splitSignals,3),(floor(size(splitSignals,1)/(frequency*window))-1)*window/shift+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(:,i,j))),[],[],NFFT,frequency);
+        end
+        tempEEG2(i,:,:)=tempEEG;
+    end
+    trainingDataEEG=permute(tempEEG2,[2 1 3]);
+    
+    parfor i=33:size(splitSignals,2)-3
+        tempEMG=zeros([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;
+    end
+    trainingDataEMG=permute(tempEMG2,[2 1 3]);
+end
\ No newline at end of file
diff --git a/kfoldCV.m b/kfoldCV.m
new file mode 100644
index 0000000..9deb989
--- /dev/null
+++ b/kfoldCV.m
@@ -0,0 +1,32 @@
+function [model] = kfoldCV(classification,trainingData,k,cExpMax)
+
+    randMap=randperm(size(trainingData,1));
+    leaveOut=trainingData(mod(randMap,k+1)==0,:,:);
+    leaveClasses=classification(mod(randMap,k+1)==0);
+    remaining=trainingData(mod(randMap,k+1)~=0,:,:);
+    remainingClasses=classification(mod(randMap,k+1)~=0);
+    cvAccurancy=zeros(2*cExpMax+1,1);
+    for cExp=-cExpMax:1:cExpMax
+        c=2^cExp;
+
+        randomMapping=randperm(size(remaining,1));
+        accurancy=zeros(k,3);
+
+        parfor i=1:k
+            trainData=remaining(mod(randomMapping,k)+1~=i,:,:);
+            testData=remaining(mod(randomMapping,k)+1==i,:,:);
+            trainClasses=remainingClasses(mod(randomMapping,k)+1~=i);
+            testClasses=remainingClasses(mod(randomMapping,k)+1==i);
+            [trainClasses,trainData]=balanceClasses(trainClasses,trainData,0.00001);
+            model=svmtrain(trainClasses,trainData(:,:),sprintf('-t 0 -c %f -q',c));
+            [~, accurancy(i,:), ~]=svmpredict(testClasses, testData(:,:), model,'-q');
+        end
+
+        cvAccurancy=mean(accurancy(:,1));
+    end
+    [~,i]=max(cvAccurancy);
+    bestC=2^(i-cExpMax-1);
+    [balanceLeftClasses, balanceLeftOut]=balanceClasses(leaveClasses,leaveOut,0.00001);
+    model=svmtrain(balanceLeftClasses, balanceLeftOut(:,:),sprintf('-t 0 -c %f -q',bestC));
+
+end
\ No newline at end of file
diff --git a/myFilter.m b/myFilter.m
new file mode 100644
index 0000000..8f6c13d
--- /dev/null
+++ b/myFilter.m
@@ -0,0 +1,26 @@
+function [ filteredSig, pwMeanMove, pwMeanRest, hz ] = myFilter( splitSignals,code,NFFT )
+    %Filter around 50Hz and below 2 Hz
+    [A,B]= butter(2,[48 52]/1250,'stop');
+    [C,D]= butter(2,2/1250,'high');
+    
+    
+    filteredSig=zeros(size(splitSignals));
+    pwMeanMove=zeros(NFFT/2+1,size(splitSignals,2));
+    pwMeanRest=zeros(NFFT/2+1,size(splitSignals,2));
+
+    for i=1:size(splitSignals,2) %filter single channel
+        for j=1:size(splitSignals,3) %filter single trial
+            filteredSig(:,i,j)=filter(C,D,filter(A,B,splitSignals(:,i,j)));
+            if code(j)<=5
+                pwMeanMove(:,i)=pwMeanMove(:,i)+pwelch(filteredSig(:,i,j),[],[],NFFT,2500);
+            elseif code(j) > 5
+                pwMeanRest(:,i)=pwMeanRest(:,i)+pwelch(filteredSig(:,i,j),[],[],NFFT,2500);
+            end
+            %s=cat(1,s,size(pwelch(filteredSig(:,i,j),[],[],[],2500)));
+        end
+        pwMeanMove(:,i)=pwMeanMove(:,i) /size(find(code <=5),1);
+        pwMeanRest(:,i)=pwMeanRest(:,i) /size(find(code > 5),1);
+    end
+    [a,hz]=pwelch(filteredSig(:,1,1),[],[],NFFT,2500);
+end
+
diff --git a/plotChannelDiff.m b/plotChannelDiff.m
new file mode 100644
index 0000000..e3e0e01
--- /dev/null
+++ b/plotChannelDiff.m
@@ -0,0 +1,17 @@
+function [] = plotChannelDiff( pwMeanMove, pwMeanRest,hz, params )
+
+for i=1:size(params.ChannelNames.Value,1)
+    n=mod(i,9);
+    if n==1
+        figure
+    elseif n==0;
+        n=9;
+    end
+    subplot(3,3,n);
+    plot(hz,pwMeanMove(:,i)-pwMeanRest(:,i))
+    xlim([0 150])
+    title(params.ChannelNames.Value(i))
+end
+
+end
+
diff --git a/plotChannelProzDiff.m b/plotChannelProzDiff.m
new file mode 100644
index 0000000..b57bd79
--- /dev/null
+++ b/plotChannelProzDiff.m
@@ -0,0 +1,17 @@
+function [] = plotChannelProzDiff( pwMeanMove, pwMeanRest,hz, params )
+
+for i=1:size(params.ChannelNames.Value,1)
+    n=mod(i,9);
+    if n==1
+        figure
+    elseif n==0;
+        n=9;
+    end
+    subplot(3,3,n);
+    plot(hz,(pwMeanMove(:,i)-pwMeanRest(:,i))./pwMeanRest(:,i))
+    xlim([0 150])
+    title(params.ChannelNames.Value(i))
+end
+
+end
+
diff --git a/plotChannels.m b/plotChannels.m
new file mode 100644
index 0000000..9126ec1
--- /dev/null
+++ b/plotChannels.m
@@ -0,0 +1,17 @@
+function [] = plotChannels( pwMean, hz, params )
+
+for i=1:size(params.ChannelNames.Value,1)
+    n=mod(i,9);
+    if n==1
+        figure
+    elseif n==0;
+        n=9;
+    end
+    subplot(3,3,n);
+    plot(log(hz),log(pwMean(:,i,:)))
+    xlim([0 150])
+    title(params.ChannelNames.Value(i))
+end
+
+end
+
diff --git a/plotImgScProzDiff.m b/plotImgScProzDiff.m
new file mode 100644
index 0000000..a1e6b34
--- /dev/null
+++ b/plotImgScProzDiff.m
@@ -0,0 +1,8 @@
+function [] = plotImgScProzDiff( pwMeanMove, pwMeanRest)
+
+prozDiff=(pwMeanMove-pwMeanRest)./pwMeanRest;
+imagesc(1:1.22:100,1:32,transpose(prozDiff(1:82,1:32)))
+colorbar();
+
+end
+
diff --git a/readme.m b/readme.m
new file mode 100644
index 0000000..13fd3d6
--- /dev/null
+++ b/readme.m
@@ -0,0 +1,55 @@
+[signal, states, params] = load_bcidat('/home/jph/Uni/Masterarbeit/Block1_ReachingMovements/AO/AO_B1001/AO_B1S001R01.dat');
+params.SamplingRate.NumericValue
+plot(states.StimulusCode)
+
+% digitale Filter
+help butter
+help cheby1
+help cheby2
+
+[A,B]= butter(2,[48 52]/1250,'stop');
+
+filteredSig = filter(A,B,signal(:,14));
+
+% Frequenzspektrum
+pwelch(signal(:,14),[],[],[],2500)
+pburg
+
+%-----------
+% 1. EEG schneiden: Trials, Labels (Ruhe und Bewegung)
+% (optinal) Filter (50 Hz, highpass (0.1Hz?)
+% 2. Frequenzspektrum
+% 3. Mittelwerte Spektrum �ber Trials: Ruhe gegen Bewegung
+% auf C3, C4 konzentrieren (10-20 elektroden system nachschauen)
+
+%PSDC3=pwelch(filteredSig(:,:,1),[],[],[],2500);
+%pwelch(filteredSig(:,13,16),[],[],[],2500);
+%xlim([0 0.100])
+
+% % plot all
+% for i=1:size(filteredSig,3)
+%     if mod(i,4)==0
+%         figure
+%     end
+%     subplot(2,2,mod(i,4)+1);
+%     pwelch(filteredSig(:,13,i),[],[],[],2500);
+%     xlim([0 0.100]);
+% end
+
+%TODO
+%1. filtern in zwei for-Schleifen um sicher zu gehen
+%2. anschauen, evtl Window (1s)
+%3. Differenz (bzw. prozentualer Unterschied) Trial vs. Ruhe
+
+%Tipps
+% find, diff, xlim([0 0.1])
+
+%imagesc
+
+%libsvm
+% - svmTrain
+%trial x ch x power (1-50) | bewegung / ruhe
+% cross validation (selber impelmentieren)
+
+% Aufteilung EMG xor EEG (done)
+% danach dann Bewegungen aufteilen
\ No newline at end of file
diff --git a/stimuliMatrix.txt b/stimuliMatrix.txt
new file mode 100644
index 0000000..091d456
--- /dev/null
+++ b/stimuliMatrix.txt
@@ -0,0 +1,4 @@
+BlueGo	RedGo	GreenGo	BrownGo	PinkGo	REST	REST	REST	REST	REST	REST	REST	REST	REST	REST	%
+%	%	%	%	%	%	%	%	%	%	%	%	%	%	%	%
+sounds\BlueGo.wav	sounds\RedGo.wav	sounds\GreenGo.wav	sounds\BrownGo.wav	sounds\PinkGo.wav	sounds\Rest.wav	sounds\Rest.wav	sounds\Rest.wav	sounds\Rest.wav	sounds\Rest.wav	sounds\Rest.wav	sounds\Rest.wav	sounds\Rest.wav	sounds\Rest.wav	sounds\Rest.wav	%
+6000ms	6000ms	6000ms	6000ms	6000ms	2000ms	2100ms	2200ms	2300ms	2400ms	2500ms	2600ms	2700ms	2800ms	2900ms	2000ms
diff --git a/svm.m b/svm.m
new file mode 100644
index 0000000..f30be1f
--- /dev/null
+++ b/svm.m
@@ -0,0 +1,36 @@
+%params
+NFFT=2048;
+cutWindow=2;
+window=0.2;
+shift=0.05;
+k=10;
+rounds=1;
+maxExpC=5;
+
+%svm
+[signal, states, params] = load_bcidat('/home/jph/Uni/masterarbeit/Block1_ReachingMovements/AO/AO_B1001/AO_B1S001R01.dat');
+% [ splitSignals, code ] = cutTrials( states, signal );
+[ splitSignals, code ] = cutWindows( states, signal, params, cutWindow );
+
+%trainingData has format trial x ch x powerSpectrum
+[trainingDataEEG,trainingDataEMG]=generateTrainingData(splitSignals,NFFT,window,shift,params);
+
+classification=code;%double(code<=5);
+
+
+accurancy=zeros(rounds,3);
+means=zeros(rounds,1);
+for i=1:rounds
+    randMap=randperm(size(trainingDataEEG,1));
+    leaveOut=trainingDataEEG(mod(randMap,k+2)==0,:,:);
+    leaveClasses=classification(mod(randMap,k+2)==0);
+    remaining=trainingDataEEG(mod(randMap,k+2)~=0,:,:);
+    remainingClasses=classification(mod(randMap,k+2)~=0);
+
+    model=kfoldCV(remainingClasses,remaining,k,maxExpC);
+
+    [predictions,accurancy(i,:),pvalues]=svmpredict(leaveClasses,leaveOut(:,:),model);
+    means(i)=mean(leaveClasses-predictions);
+end
+m=mean(accurancy(:,1))
+mm=mean(means)
\ No newline at end of file
diff --git a/waveformLength.m b/waveformLength.m
new file mode 100644
index 0000000..cfc0298
--- /dev/null
+++ b/waveformLength.m
@@ -0,0 +1,12 @@
+function [WL]=waveformLength(signal,frequency,window,windowShift)
+%waveformLength computes waveformLength of signal with given frequency, window and
+%shift
+    signalShift=frequency*windowShift;
+    signalWindow=frequency*window;
+    shiftProp=window/windowShift;
+    WL=zeros((floor(size(signal,1)/signalWindow)-1)*shiftProp+1,1);
+    parfor j=1:(size(signal,1)/signalWindow-1)*shiftProp+1
+        WL(j)=sum(diff(signal((j-1)*signalShift+1:(j-1)*signalShift+signalWindow)));
+    end
+end
+