// konstruktor MOVING_AVG, SAV_GOL, LMS LOW, HIGH
public ECG_Baseline_Params(Filtr_Method method, Filtr_Type type, int windowSize, string analysisName)
{
this.Method = method;
this.AnalysisName = analysisName;
this.Type = type;
if (type == Filtr_Type.LOWPASS)
this.WindowSizeLow = windowSize;
else if (type == Filtr_Type.HIGHPASS)
this.WindowSizeHigh = windowSize;
}
//konstruktor BUTTERWORTH BAND
public ECG_Baseline_Params(Filtr_Method method, Filtr_Type type, int orderLow, int orderHigh, double fcLow, double fcHigh, string analysisName)
{
this.Method = method;
this.AnalysisName = analysisName;
this.Type = type;
this.FcLow = fcLow;
this.FcHigh = fcHigh;
this.OrderLow = orderLow;
this.OrderHigh = orderHigh;
}
//konstruktor BUTTERWORTH LOW, HIGH
public ECG_Baseline_Params(Filtr_Method method, Filtr_Type type, int order, double fc, string analysisName)
{
this.Method = method;
this.AnalysisName = analysisName;
this.Type = type;
if (type == Filtr_Type.LOWPASS)
this.FcLow = fc;
else if (type == Filtr_Type.HIGHPASS)
this.FcHigh = fc;
if (type == Filtr_Type.LOWPASS)
this.OrderLow = order;
else if (type == Filtr_Type.HIGHPASS)
this.OrderHigh = order;
}
public Vector<double> butterworth(Vector<double> signal, double fs, double fc, int order, Filtr_Type type)
{
//TODO: Comments
int L = signal.Count;
int DCGain = 1;
double[] DoubleArray = new double[L];
DoubleArray = signal.ToArray();
Complex[] ComplexArray = new Complex[L];
for (int i = 0; i < L; i++)
{
ComplexArray[i] = new Complex(DoubleArray[i], 0);
}
Fourier.Forward(ComplexArray, FourierOptions.Matlab);
double binWidth, binFreq, gain;
if (fc > 0)
{
binWidth = fs / L;
for(int i = 0; i < (L/2); i++)
{
binFreq = binWidth * (i+1);
gain = DCGain / Math.Sqrt(1 + Math.Pow(binFreq/fc, 2.0 * order));
if (type == Filtr_Type.HIGHPASS)
{
gain = 1 - gain;
}
ComplexArray[i] *= gain;
ComplexArray[L - 1 - i] *= gain;
}
}
Fourier.Inverse(ComplexArray, FourierOptions.Matlab);
for (int i = 0; i < L; i++)
{
DoubleArray[i] = ComplexArray[i].Real;
}
Vector<double> output_signal = Vector<double>.Build.DenseOfArray(DoubleArray);
return output_signal;
}
// konstruktor MOVING_AVG, SAV_GOL, LMS BAND
public ECG_Baseline_Params(Filtr_Method method, Filtr_Type type, int windowSizeLow, int windowSizeHigh, string analysisName)
{
this.Method = method;
this.AnalysisName = analysisName;
this.Type = type;
this.WindowSizeLow = windowSizeLow;
this.WindowSizeHigh = windowSizeHigh;
}
public Vector<double> savitzky_golay(Vector<double> signal, int window_size, Filtr_Type type)
{
//TODO: Comments
int signal_size = signal.Count;
if ((window_size % 2) == 0)
{
window_size = window_size + 1;
}
Vector<double> signal_extension_front = Vector<double>.Build.Dense((window_size / 2), signal[0]);
Vector<double> signal_extension_back = Vector<double>.Build.Dense((window_size / 2), signal[signal_size - 1]);
int signal_extension_size = signal_extension_front.Count;
Vector<double> signal_extended = Vector<double>.Build.Dense(2*signal_extension_size + signal_size, 0); //przygotowanie zmiennej, w której znajdzie się sygnał przygotowany do filtracji
Vector<double> signal_filtered = Vector<double>.Build.Dense(signal_size, 0); //zmienna na sygnał przefiltrowany
for (int i = 0; i < signal_extension_size; i++)
{
signal_extended[i] = signal_extension_front[i]; //powielenie piewszej próbki sygnału wejściowego
}
for (int i = 0; i < signal_size; i++)
{
signal_extended[i + signal_extension_size] = signal[i]; //powielenie piewszej próbki sygnału wejściowego
}
for (int i = 0; i < signal_extension_size; i++)
{
signal_extended[i + signal_extension_size + signal_size] = signal_extension_back[i]; //powielenie piewszej próbki sygnału wejściowego
}
Vector<double> samples = Vector<double>.Build.Dense(signal_extended.Count, 0);
for (int i = 0; i < samples.Count; i++)
{
samples[i] = i + 1;
}
double[] coeff = new double[3];
Vector<double> samples_destination = Vector<double>.Build.Dense(window_size, 0);
Vector<double> signal_extended_destination = Vector<double>.Build.Dense(window_size, 0);
double[] output_signal_table = new double[signal_size];
for (int i = 0; i < signal_size; i++)
{
samples.CopySubVectorTo(samples_destination, i, 0, window_size);
signal_extended.CopySubVectorTo(signal_extended_destination, i, 0, window_size);
coeff = Fit.Polynomial(samples_destination.ToArray(), signal_extended_destination.ToArray(), 3, DirectRegressionMethod.NormalEquations);
output_signal_table[i] = coeff[3] * Math.Pow(samples[i + (window_size / 2)], 3) + coeff[2] * Math.Pow(samples[i + (window_size / 2)], 2) + coeff[1] * samples[i + (window_size / 2)] + coeff[0];
}
Vector<double> output_signal = Vector<double>.Build.DenseOfArray(output_signal_table);
if (type == Filtr_Type.HIGHPASS)
{
return signal - output_signal;
}
return output_signal;
}
public Vector<double> moving_average(Vector<double> signal, int window_size, Filtr_Type type)
{
int signal_size = signal.Count; //rozmiar sygału wejściowego
Vector<double> signal_extension = Vector<double>.Build.Dense(window_size - 1, signal[0]); //uwzględnienie warunków początkowych filtracji
Vector<double> signal_extended = Vector<double>.Build.Dense(signal_extension.Count + signal_size, 0); //przygotowanie zmiennej, w której znajdzie się sygnał przygotowany do filtracji
Vector<double> signal_filtered = Vector<double>.Build.Dense(signal_size, 0); //zmienna na sygnał przefiltrowany
for (int i = 0; i < signal_extension.Count; i++)
{
signal_extended[i] = signal_extension[i]; //powielenie piewszej próbki sygnału wejściowego
}
for (int i = 0; i < signal_size; i++)
{
signal_extended[i + window_size - 1] = signal[i]; //przygotowanie sygnału do filtracji
}
double sum = 0; //zmienna pomocnicza
for (int i = 0; i < signal_size; i++)
{
sum = 0;
for (int j = 0; j < window_size; j++)
{
sum = sum + signal_extended[i + j]; //sumowanie kolejnych próbek sygnału
}
signal_filtered[i] = sum / window_size; //obliczenie średniej
}
if(type == Filtr_Type.HIGHPASS)
{
return signal - signal_filtered;
}
return signal_filtered; //sygnał przefiltrowany
}
