private void OnLoaded(object sender, RoutedEventArgs routedEventArgs)
{
var pointsPerSlice = 100;
var sliceCount = 20;
var logBase = 10;
var slicePositions = new double[sliceCount];
for (int i = 0; i < sliceCount; ++i)
{
slicePositions[i] = Math.Pow(logBase, i);
}
var dataSeries = new WaterfallDataSeries3D <double>(pointsPerSlice, slicePositions)
{
SeriesName = "Waterfall"
};
dataSeries.StartX = 10;
dataSeries.StepX = 1;
_transform.init((uint)Math.Log(pointsPerSlice, 2));
FillDataSeries(dataSeries, sliceCount, pointsPerSlice);
WaterfallSeries.DataSeries = dataSeries;
}
public void TestFFT()
{
const int SIZE = 512;
var re = new double[SIZE];
var im = new double[SIZE];
for (int i = 0; i < SIZE; i++)
{
re[i] = Math.Sin(2 * Math.PI / SIZE * i);
im[i] = 0;
}
var fft = new FFT2();
var logN = (uint)Math.Log(SIZE, 2);
fft.init(logN);
var expected = new double[SIZE];
Array.Copy(re, expected, re.Length);
fft.run(re, im);
fft.run(re, im, true);
for (int i = 0; i < SIZE; i++)
{
Assert.AreEqual(expected[i], re[i], 0.000000001);
}
}
public void AnalyzeData()
{
print("AnalyzeData");
total_Acc = new double[raw_data.Count];
textMesh.text = "Calculating total acceleration on " + raw_data.Count.ToString() + " values";
for (int i = 0; i < raw_data.Count; i++)
{
//calculate total acceleration of the 3 axis
total_Acc[i] = Mathf.Sqrt(Mathf.Pow((raw_data.ElementAt(i).x), 2) + Mathf.Pow((raw_data.ElementAt(i).y), 2) + Mathf.Pow((raw_data.ElementAt(i).z), 2));
//print("\t" + total_Acc[i]);
}
FFT2 fft2 = new FFT2();
/**
* Initialize class to perform FFT of specified size.
*
* @param logN Log2 of FFT length. e.g. for 512 pt FFT, logN = 9.
*/
textMesh.text = "starting FFT";
fft2.init((uint)Mathf.Log(total_Acc.Length));
//create array of double for Im part-----> array should be compsed by 0
y_fft = new double[total_Acc.Length];
for (int i = 0; i < total_Acc.Length; i++)
{
y_fft[i] = 0;
}
//run the fft
fft2.run(total_Acc, y_fft);
StartCoroutine(ShowResult());
}
public VoiceActivityDetectorV2(int sampleRate)
{
this.sampleRate = sampleRate;
samplesInFrame = sampleRate / (1000 / frameSize);
fft = new FFT2();
fft.init((uint)Math.Log(samplesInFrame, 2));
Enabled = true; // default value
}
private static WaterfallDataSeries3D <double> GetWaterfallDataSeries()
{
var pointsPerSlice = 100;
var sliceCount = 20;
var logBase = 10;
var slicePositions = new double[sliceCount];
for (int i = 0; i < sliceCount; ++i)
{
slicePositions[i] = i;
}
var dataSeries = new WaterfallDataSeries3D <double>(pointsPerSlice, slicePositions);
dataSeries.StartX = 10;
dataSeries.StepX = 1;
_transform.init((uint)Math.Log(pointsPerSlice, 2));
var count = pointsPerSlice * 2;
var re = new double[count];
var im = new double[count];
for (int sliceIndex = 0; sliceIndex < sliceCount; ++sliceIndex)
{
for (var i = 0; i < count; i++)
{
re[i] = 2.0 * Math.Sin(2 * Math.PI * i / 20) +
5 * Math.Sin(2 * Math.PI * i / 10) +
2.0 * _random.NextDouble();
im[i] = -10;
}
_transform.run(re, im);
var scaleCoef = Math.Pow(1.5, sliceIndex * 0.3) / Math.Pow(1.5, sliceCount * 0.3);
for (var pointIndex = 0; pointIndex < pointsPerSlice; pointIndex++)
{
var mag = Math.Sqrt(re[pointIndex] * re[pointIndex] + im[pointIndex] * im[pointIndex]);
var yVal = _random.Next(10, 20) * Math.Log10(mag / pointsPerSlice);
yVal = (yVal <-25 || yVal> -5)
? (yVal < -25) ? -25 : _random.Next(-6, -3)
: yVal;
dataSeries[sliceIndex, pointIndex] = -yVal * scaleCoef;
}
}
return(dataSeries);
}
private void LoadFile(string path, ArrayViewer waveViewer, ArrayViewer fftViewer)
{
FileStream fileStream = new FileStream(path, FileMode.Open);
BinaryReader binaryReader = new BinaryReader(fileStream);
short[] wave = new short[binaryReader.BaseStream.Length / 2];
int i = 0;
while (binaryReader.BaseStream.Position < binaryReader.BaseStream.Length)
{
uint a = binaryReader.ReadByte();
uint b = binaryReader.ReadByte();
wave[i] = (short)((a << 8) | b);
i++;
}
fileStream.Close();
fileStream.Dispose();
waveViewer.DrawWave(wave);
Application.DoEvents();
FFT2 fft2 = new FFT2();
fft2.init((uint)Math.Log((double)wave.Length, 2.0));
double[] re = new double[wave.Length];
double[] img = new double[wave.Length];
for (int j = 0; j < wave.Length; j++)
{
re[j] = wave[j];
img[j] = 0.0F;
}
fft2.run(re, img);
double[] modulo = new double[re.Length / 2];
for (int j = 0; j < modulo.Length; j++)
{
modulo[j] = Math.Sqrt((re[j] * re[j]) + (img[j] * img[j]));
}
int number = (int)(2048.0F*(float) modulo.Length/8000.0F);
fftViewer.DrawWave(modulo.Take(number).ToArray());
}
private void OnLoaded(object sender, RoutedEventArgs routedEventArgs)
{
// Initialize FFT
_transform.init((uint)Math.Log(_pointsPerSlice, 2));
_transformSize = _pointsPerSlice * 2;
_real = new double[_transformSize];
_imaginary = new double[_transformSize];
// Initialize WaterfallDataSeries3D
_waterfallDataSeries = new WaterfallDataSeries3D <double>(_pointsPerSlice, _maxSliceCount);
_waterfallDataSeries.StartX = 10;
_waterfallDataSeries.StepX = 1;
_waterfallDataSeries.StartZ = 25;
_waterfallDataSeries.StepZ = 10;
WaterfallSeries.DataSeries = _waterfallDataSeries;
MeshSeries.DataSeries = (UniformGridDataSeries3D <double, double, double>)_waterfallDataSeries;
}
//public int numPartitions = 10000;
//public float overlapPercent = 0.5f;
//public float threshold = 1 - 0.75f; //larger float values are more strict
//public float beatDetectionOverlapPercent = 0.5f;
/// <summary>
/// Processes raw audio data to find average energy for overlapping partitions.
/// </summary>
/// <returns>The FFT data array.</returns>
/// <param name="clip">Audio clip to process.</param>
/// <param name="numPartitions">Number of pieces to split the song into for analysis</param>
/// <param name="overlapPercent">The percentage which the partitions overlap each other</param>
public double[] ProcessAudio(AudioClip clip, int numPartitions, float overlapPercent)
{
_averages = new double[(int)(numPartitions / overlapPercent) - 1];
int samplesPerPartition = (int)(clip.samples / numPartitions);
int numDivisions = (int)(numPartitions / overlapPercent) - 1;
//Because the partitions overlap, the number of iterations is the number of partitions multiplied by the inverse of the overlap percent
for (int i = 0; i < numDivisions; i++)
{
float[] samples = new float[samplesPerPartition];
int input = i * ((int)(samples.Length * overlapPercent)); //the offset to start getting song data increases by overlapPercent as i is incremented
clip.GetData(samples, input);
//the raw partition data is run through the Blackman-Harris windowing function
for (int n = 0; n < samples.Length; n++)
{
samples [n] *= _a0 - _a1 * Mathf.Cos((2 * Mathf.PI * n) / samples.Length - 1) + _a2 * Mathf.Cos((4 * Mathf.PI * n) / samples.Length - 1) - _a3 * Mathf.Cos((6 * Mathf.PI * n) / samples.Length - 1);
}
FFT2 FFT = new FFT2();
FFT.init((uint)Mathf.Log(samplesPerPartition, 2));
//our array of floats is converted to an array of doubles for use in the FFT function
double[] double_samples = samples.ToList().ConvertAll <double> (new System.Converter <float, double> (F2D)).ToArray();
//runs our sample data through a Fast Fourier Transform to convert it to the frequency domain
FFT.run(double_samples, new double[samples.Length], false);
//gets the average value for this partition and adds it to an array.
//when all of the partitions are completed, averages will contain data for the entire song
double avg = double_samples.Average();
_averages[i] = avg;
}
return(_averages);
}
//public int numPartitions = 10000;
//public float overlapPercent = 0.5f;
//public float threshold = 1 - 0.75f; //larger float values are more strict
//public float beatDetectionOverlapPercent = 0.5f;
/// <summary>
/// Processes raw audio data to find average energy for overlapping partitions.
/// </summary>
/// <returns>The FFT data array.</returns>
/// <param name="clip">Audio clip to process.</param>
/// <param name="numPartitions">Number of pieces to split the song into for analysis</param>
/// <param name="overlapPercent">The percentage which the partitions overlap each other</param>
public double[] ProcessAudio(AudioClip clip, int numPartitions, float overlapPercent)
{
_averages = new double[(int)(numPartitions / overlapPercent) - 1];
int samplesPerPartition = (int)(clip.samples / numPartitions);
int numDivisions = (int)(numPartitions / overlapPercent) - 1;
//Because the partitions overlap, the number of iterations is the number of partitions multiplied by the inverse of the overlap percent
for (int i = 0; i < numDivisions; i++) {
float[] samples = new float[samplesPerPartition];
int input = i * ((int) (samples.Length * overlapPercent)); //the offset to start getting song data increases by overlapPercent as i is incremented
clip.GetData(samples, input);
//the raw partition data is run through the Blackman-Harris windowing function
for (int n = 0; n < samples.Length; n++) {
samples [n] *= _a0 - _a1 * Mathf.Cos ((2 * Mathf.PI * n) / samples.Length - 1) + _a2 * Mathf.Cos ((4 * Mathf.PI * n) / samples.Length - 1) - _a3 * Mathf.Cos ((6 * Mathf.PI * n) / samples.Length - 1);
}
FFT2 FFT = new FFT2 ();
FFT.init ((uint)Mathf.Log(samplesPerPartition,2));
//our array of floats is converted to an array of doubles for use in the FFT function
double[] double_samples = samples.ToList ().ConvertAll<double> (new System.Converter<float, double> (F2D)).ToArray ();
//runs our sample data through a Fast Fourier Transform to convert it to the frequency domain
FFT.run (double_samples, new double[samples.Length], false);
//gets the average value for this partition and adds it to an array.
//when all of the partitions are completed, averages will contain data for the entire song
double avg = double_samples.Average ();
_averages[i] = avg;
}
return _averages;
}
double[] DoFFT(double[] values) {
double[] real = new double[values.Length];
double[] imaginary = new double[values.Length];
for (int i = 0; i < values.Length; i++) {
imaginary[i] = 0f;
real[i] = values[i];
}
FFT2 fft = new FFT2();
fft.init((uint)Mathf.Log ((float)values.Length));
fft.run (real, imaginary, false);
return real;
}