public void GetForwardPrediction_ZeroInput_ReturnsCorrectPrediction()
{
const int coefNumber = 4;
const int historyLength = 512;
const int numberOfSamplesToCheck = 10;
var inputAudio =
new double[historyLength + numberOfSamplesToCheck];
for (var i = 0; i < inputAudio.Length; i++)
{
inputAudio[i] = 0;
}
var fba = new FastBurgAlgorithm128(inputAudio);
for (var index = historyLength;
index < historyLength + numberOfSamplesToCheck;
index++)
{
fba.Train(index, coefNumber, historyLength);
var forwardPrediction = fba.GetForwardPrediction();
Assert.AreEqual(
inputAudio[index],
forwardPrediction,
0.0000001);
}
}
public void GetBackwardPrediction_SinInput_ReturnsCorrectPrediction()
{
const int coefNumber = 4;
const int historyLength = 512;
const int numberOfSamplesToCheck = 10;
var inputAudio =
new double[historyLength + numberOfSamplesToCheck];
for (var i = 0; i < inputAudio.Length; i++)
{
inputAudio[i] = Math.Sin(
2 * Math.PI * i / (historyLength / 5.2));
}
var fba = new FastBurgAlgorithm128(inputAudio);
for (var index = historyLength + 1;
index < historyLength + numberOfSamplesToCheck;
index++)
{
fba.Train(index, coefNumber, historyLength);
var backwardPrediction = fba.GetBackwardPrediction();
Assert.AreEqual(
inputAudio[index - historyLength - 1],
backwardPrediction,
0.0000001);
}
}
public void Train_SinInput_ReturnsCorrectPredictionAndReflectionCoefficients()
{
/* Coefficients for comparision are taken from GNU Octave arburg() function
* t = [0:2000]
* x = sin( 2 * pi() * t / (512 / 5.2))
* output_precision(16)
* [a, v, k] = arburg(x(1:512),4)
*/
const int coefNumber = 4;
const int historyLength = 512;
const int numberOfSamplesToCheck = 1;
const double accuracy = 0.0000000000001;
var inputAudio =
new double[historyLength + numberOfSamplesToCheck];
for (var i = 0; i < inputAudio.Length; i++)
{
inputAudio[i] = Math.Sin(
2 * Math.PI * i / (historyLength / 5.2));
}
var fba = new FastBurgAlgorithm128(inputAudio);
fba.Train(historyLength, coefNumber, historyLength);
var predictionCoefs = fba.GetPredictionCoefs();
Assert.AreEqual(
1,
(double)predictionCoefs[0],
accuracy);
Assert.AreEqual(
-3.991510267867756,
(double)predictionCoefs[1],
accuracy);
Assert.AreEqual(
5.983035128379795,
(double)predictionCoefs[2],
accuracy);
Assert.AreEqual(
-3.991503459864878,
(double)predictionCoefs[3],
accuracy);
Assert.AreEqual(
0.9999965889050035,
(double)predictionCoefs[4],
accuracy);
var reflectionCoefs = fba.GetReflectionCoefs();
Assert.AreEqual(
-0.9979213453536945,
(double)reflectionCoefs[0],
accuracy);
Assert.AreEqual(
0.9999990984096440,
(double)reflectionCoefs[1],
accuracy);
Assert.AreEqual(
-0.9978363901155060,
(double)reflectionCoefs[2],
accuracy);
Assert.AreEqual(
0.9999965889050035,
(double)reflectionCoefs[3],
accuracy);
}
