cvec prePost (cvec data_time, int pre, int post){
data_time.ins(0, data_time.right(pre));
data_time.ins(data_time.length(), data_time.mid(pre,post));
return(data_time);
}
/** Return the channel FIR estimation based on trainSeq and fill phiHat, Ahat, delay, sigmaSqrNoise
*
* @pre:
* - dataC: cvec of the received data
* - trainC: pointer to array of the training sequence
* !!! length assume to be < than dataC.length()!!!
* - Post and Pre: FIR filter channel number of past and future elements
* - phiHat: pointer to a double with phase estimation
* - delay: pointer to a int with synchornization point
* - sigmaSqrNoise: pointer to a double with FIR estimator variance
*
* @post:
* - Estimators are now filled!
* - returned the channel FIR estimation
*/
cvec synchCatchChannel(cvec dataC, cvec trainCUp, int Post, int Pre , double *phiHat, double *AHat, int *delay, double *sigmaSqrNoise ){
if(dataC.length()<trainCUp.length()){
cerr << "The training sequence is too big compare to datas!\n";
exit(1);
}
if(Pre<0||Post<0){
cerr << "Not possible to compute filter!\n";
exit(1);
}
//std::cout<<"Data="<<dataC<<"\n";
//std::cout<<"Train="<<trainCUp<<"\n";
//Computes crosscorrelation:
cvec crossCorr=itpp::xcorr(dataC, trainCUp);
//std::cout<<"Cross="<<crossCorr<<"\n";
//Search for the maximum
double max=0.0;
int index=0, count=0;
double tmp[crossCorr.length()-dataC.length()+1];
DispVal(crossCorr.length()-dataC.length()+1);
for(int i=dataC.length()-1;i<crossCorr.length();i++){
tmp[count]=abs(crossCorr[i]);
//DispVal(tmp[count]);
//DispVal(count);
if (tmp[count]> max && count<(crossCorr.length()-dataC.length()+1)/2 ){
max=tmp[count];
index=i;
DispVal(index-dataC.length()+1);
}
count++;
}
index=index-dataC.length()+1;
DispVal(index);
*delay=index;
//Computes Estimate of phase at the peak:
*phiHat=arg(dataC(index));
// Save data to file
std::ofstream ofs( "xcorr.dat" , std::ifstream::out );
ofs.write((char * ) tmp, (crossCorr.length()-dataC.length()+1)*sizeof(double));
ofs.close();
//std::cout<<"tsamp="<<index<<"\n";
//Removes received train from data
cvec trainRecC(trainCUp.length());
int i=index;
for(int count=0;count<trainCUp.length();count++){
trainRecC.set(count,dataC[i+count]);
//DispVal(dataC[i]);
//DispVal(count);
}
//std::complex< double > meanTrain=itpp::mean(trainRecC);
//DispVal(meanTrain);
//trainRecC=trainRecC-meanTrain;
//std::cout<<"TrainRec"<<trainRecC<<"\n";
//Channel estimator using training sequence:
//autoCorr of known training sequence:
cvec autoCorr=itpp::xcorr(trainCUp);
//DispVal(autoCorr.length());
//DispVal(autoCorr(trainCUp.length()-1));
//std::cout<<"Auto="<<autoCorr<<"\n";
*AHat=abs(crossCorr(index+dataC.length()-1))/abs(autoCorr(trainCUp.length()-1));
//SigmaYY:
cvec aux(Pre+Post+1);
int start=trainCUp.length()-1;
for(int count=0; count<Pre+Post+1; count++){
if(count<=2*trainCUp.length()-1){
aux.set(count,autoCorr[start+count]);
}else{
aux.set(count,complex<double>(0.0,0.0));
}
}
cmat SigmaYY=toeplitz(aux);
//std::cout<<"SigmaYY=\n"<<SigmaYY<<"\n";
cvec aux2 =itpp::xcorr(trainRecC,trainCUp);
//SigmaYx:
int ini=trainCUp.length()-Pre-1;
cvec SigmaYx(Pre+Post+1);
for(int count=0; count<SigmaYx.length(); count++){
SigmaYx.set(count, aux2[ini+count]);
//std::cout<<"xcorr=\n"<<crossCorr[i]<<"\n";
}
//std::cout<<"SigmaYx=\n"<<SigmaYx<<"\n";
//Estimate of the channel:
cmat invSigmaYY=itpp::inv(SigmaYY);
cvec theta_est=itpp::operator*(invSigmaYY,SigmaYx);
//std::cout<<"invSigmaYY=\n"<<invSigmaYY<<"\n";
// std::cout<<"Channel="<<theta_est<<"\n";
//Compute the variance of the estimation:
trainCUp.ins(trainCUp.length(),complex<double>(0,0));
cvec trainCEst=filter( theta_est, 1, trainCUp);
//std::cout<<"trainCEst="<<trainCEst<<"\n";
trainCEst.del(0);
cvec crossEst=itpp::xcorr(trainRecC-trainCEst);
//std::cout<<"crossEst="<<crossEst<<"\n";
*sigmaSqrNoise=abs(crossEst(trainRecC.length()-1));
return theta_est;
}
