public float[][] CreateSpectrogram(AudioSamples audioSamples, int overlap, int wdftSize)
{
float[] window = new DefaultSpectrogramConfig().Window.GetWindow(wdftSize);
float[] samples = audioSamples.Samples;
int width = (samples.Length - wdftSize) / overlap;
float[][] frames = new float[width][];
for (int i = 0; i < width; i++)
{
float[] complexSignal = fftService.FFTForward(samples, i * overlap, wdftSize, window);
float[] band = new float[(wdftSize / 2) + 1];
for (int j = 0; j < (wdftSize / 2) + 1; j++)
{
double re = complexSignal[2 * j];
double img = complexSignal[(2 * j) + 1];
re /= (float)wdftSize / 2;
img /= (float)wdftSize / 2;
band[j] = (float)((re * re) + (img * img));
}
frames[i] = band;
}
return frames;
}
public LomontFFT()
{
A = 0;
B = 1;
var config = new DefaultSpectrogramConfig();
Initialize(config.WdftSize);
}
public DefaultFingerprintConfiguration()
{
SpectrogramConfig = new DefaultSpectrogramConfig();
HashingConfig = new DefaultHashingConfig();
TopWavelets = 200;
SampleRate = 5512;
NormalizeSignal = false;
Clusters = Enumerable.Empty<string>();
}
public void CreateLogSpectrogramFromSamplesLessThanFourierTransformWindowLength()
{
var configuration = new DefaultSpectrogramConfig();
var samples = TestUtilities.GenerateRandomAudioSamples(configuration.WdftSize - 1);
var result = spectrumService.CreateLogSpectrogram(samples, configuration);
Assert.AreEqual(0, result.Count);
}
public DefaultFingerprintConfiguration()
{
SpectrogramConfig = new DefaultSpectrogramConfig();
HashingConfig = new DefaultHashingConfig();
TopWavelets = 200;
SampleRate = 5512;
NormalizeSignal = false;
Clusters = Enumerable.Empty <string>();
}
public DefaultFingerprintConfiguration()
{
SpectrogramConfig = new DefaultSpectrogramConfig();
HashingConfig = new DefaultHashingConfig();
TopWavelets = 200;
SampleRate = 5512;
HaarWaveletNorm = System.Math.Sqrt(2);
Clusters = Enumerable.Empty <string>();
}
public DefaultFingerprintConfiguration()
{
SpectrogramConfig = new DefaultSpectrogramConfig();
HashingConfig = new DefaultHashingConfig();
TopWavelets = 200;
SampleRate = 5512;
HaarWaveletNorm = System.Math.Sqrt(2);
FingerprintLengthInSeconds = (double)SamplesPerFingerprint / SampleRate;
OriginalPointSaveTransform = null;
}
public void CutLogarithmizedSpectrumOfJustOneFingerprintTest()
{
var stride = new StaticStride(0, 0);
var configuration = new DefaultSpectrogramConfig { Stride = stride };
int logSpectrumLength = configuration.ImageLength; // 128
var logSpectrum = GetLogSpectrum(logSpectrumLength);
var cutLogarithmizedSpectrum = spectrumService.CutLogarithmizedSpectrum(logSpectrum, SampleRate, configuration);
Assert.AreEqual(1, cutLogarithmizedSpectrum.Count);
}
public void CreateLogSpectrogramFromMinimalSamplesLengthTest()
{
var configuration = new DefaultSpectrogramConfig { NormalizeSignal = false };
var samples = TestUtilities.GenerateRandomAudioSamples(new DefaultFingerprintConfiguration().SamplesPerFingerprint + configuration.WdftSize); // 8192 + 2048
SetupFftService(configuration, samples);
var result = spectrumService.CreateLogSpectrogram(samples, configuration);
logUtility.Verify(utility => utility.GenerateLogFrequenciesRanges(SampleRate, configuration), Times.Once());
Assert.AreEqual(1, result.Count);
Assert.AreEqual(configuration.ImageLength, result[0].Image.Length);
}
public DefaultFingerprintConfiguration()
{
SpectrogramConfig = new DefaultSpectrogramConfig();
HashingConfig = new DefaultHashingConfig();
TopWavelets = 200;
SampleRate = 5512;
HaarWaveletNorm = Math.Sqrt(2);
FingerprintLengthInSeconds = (double)SamplesPerFingerprint / SampleRate;
OriginalPointSaveTransform = (_ => Array.Empty <byte>());
GaussianBlurConfiguration = GaussianBlurConfiguration.None;
FrameNormalizationTransform = new LogSpectrumNormalization();
}
public void CreateLogSpectrogramTest()
{
var configuration = new DefaultSpectrogramConfig { ImageLength = 2048 };
var samples = TestUtilities.GenerateRandomAudioSamples((configuration.Overlap * configuration.WdftSize) + configuration.WdftSize); // 64 * 2048
SetupFftService(configuration, samples);
var result = spectrumService.CreateLogSpectrogram(samples, configuration);
logUtility.Verify(utility => utility.GenerateLogFrequenciesRanges(SampleRate, configuration), Times.Once());
Assert.AreEqual(1, result.Count);
Assert.AreEqual(configuration.WdftSize, result[0].Image.Length);
Assert.AreEqual(32, result[0].Image[0].Length);
}
public void CutLogarithmizedSpectrumTest()
{
var stride = new StaticStride(0, 0);
var configuration = new DefaultSpectrogramConfig { Stride = stride };
const int LogSpectrumLength = 1024;
var logSpectrum = GetLogSpectrum(LogSpectrumLength);
var cutLogarithmizedSpectrum = spectrumService.CutLogarithmizedSpectrum(logSpectrum, SampleRate, configuration);
Assert.AreEqual(8, cutLogarithmizedSpectrum.Count);
double lengthOfOneFingerprint = (double)configuration.ImageLength * configuration.Overlap / SampleRate;
for (int i = 0; i < cutLogarithmizedSpectrum.Count; i++)
{
Assert.IsTrue(
System.Math.Abs(cutLogarithmizedSpectrum[i].StartsAt - (i * lengthOfOneFingerprint)) < Epsilon);
}
}
public void GenerateLogFrequenciesRangesTest()
{
var defaultConfig = new DefaultSpectrogramConfig { UseDynamicLogBase = false, LogBase = 10 };
float[] logSpacedFrequencies = new[] // generated in matlab with logspace(2.50242712, 3.3010299957, 33)
{
318.00f, 336.81f, 356.73f, 377.83f, 400.18f, 423.85f, 448.92f, 475.47f, 503.59f, 533.38f, 564.92f,
598.34f, 633.73f, 671.21f, 710.91f, 752.96f, 797.50f, 844.67f, 894.63f, 947.54f, 1003.58f, 1062.94f,
1125.81f, 1192.40f, 1262.93f, 1337.63f, 1416.75f, 1500.54f, 1589.30f, 1683.30f, 1782.86f, 1888.31f,
2000f
};
int[] indexes = logUtility.GenerateLogFrequenciesRanges(defaultFingerprintConfiguration.SampleRate, defaultConfig);
for (int i = 0; i < logSpacedFrequencies.Length; i++)
{
var logSpacedFrequency = logSpacedFrequencies[i];
int index = logUtility.FrequencyToSpectrumIndex(logSpacedFrequency, defaultFingerprintConfiguration.SampleRate, defaultConfig.WdftSize);
Assert.AreEqual(index, indexes[i]);
}
}
public void CutLogarithmizedSpectrumWithSpectrumWhichIsLessThanMinimalLengthOfOneFingerprintTest()
{
var stride = new StaticStride(0, 0);
var config = new DefaultSpectrogramConfig { Stride = stride };
int logSpectrumLength = config.ImageLength - 1;
var logSpectrum = GetLogSpectrum(logSpectrumLength);
var cutLogarithmizedSpectrum = spectrumService.CutLogarithmizedSpectrum(logSpectrum, SampleRate, config);
Assert.AreEqual(0, cutLogarithmizedSpectrum.Count);
}
public void ShouldCreateCorrectNumberOfSubFingerprints()
{
var configuration = new DefaultSpectrogramConfig { Stride = new StaticStride(0) };
var tenMinutes = 10 * 60;
var samples = TestUtilities.GenerateRandomAudioSamples(tenMinutes * SampleRate);
SetupFftService(configuration, samples);
var result = spectrumService.CreateLogSpectrogram(samples, configuration);
Assert.AreEqual((tenMinutes * SampleRate) / (configuration.ImageLength * configuration.Overlap), result.Count);
}
private void SetupFftService(DefaultSpectrogramConfig configuration, AudioSamples samples)
{
logUtility.Setup(utility => utility.GenerateLogFrequenciesRanges(SampleRate, configuration))
.Returns(new[]
{
118, 125, 133, 141, 149, 158, 167, 177, 187, 198, 210, 223, 236, 250, 264, 280, 297, 314,
333, 352, 373, 395, 419, 443, 470, 497, 527, 558, 591, 626, 663, 702, 744,
});
fftService.Setup(service => service.FFTForward(samples.Samples, It.IsAny<int>(), configuration.WdftSize, It.IsAny<float[]>()))
.Returns(TestUtilities.GenerateRandomFloatArray(2048));
}
public void CutLogarithmizedSpectrumWithDefaultStride()
{
var config = new DefaultSpectrogramConfig();
int logSpectrumlength = config.ImageLength * 10;
var logSpectrum = GetLogSpectrum(logSpectrumlength);
var cutLogarithmizedSpectrum = spectrumService.CutLogarithmizedSpectrum(logSpectrum, SampleRate, config);
// Default stride between 2 consecutive images is 1536, but because of rounding issues and the fact
// that minimal step is 11.6 ms, timestamp is roughly .37155 sec
const double TimestampOfFingerprints = (double)1536 / SampleRate;
Assert.AreEqual(49, cutLogarithmizedSpectrum.Count);
for (int i = 0; i < cutLogarithmizedSpectrum.Count; i++)
{
Assert.IsTrue(System.Math.Abs(cutLogarithmizedSpectrum[i].StartsAt - (i * TimestampOfFingerprints)) < Epsilon);
}
}
public void CutLogarithmizedSpectrumWithAnIncrementalStaticStride()
{
var stride = new IncrementalStaticStride(new DefaultFingerprintConfiguration().SamplesPerFingerprint / 2);
var config = new DefaultSpectrogramConfig { Stride = stride };
int logSpectrumLength = (config.ImageLength * 24) + config.Overlap;
var logSpectrum = GetLogSpectrum(logSpectrumLength);
var cutLogarithmizedSpectrum = spectrumService.CutLogarithmizedSpectrum(logSpectrum, SampleRate, config);
Assert.AreEqual(48, cutLogarithmizedSpectrum.Count);
double lengthOfOneFingerprint = (double)config.ImageLength * config.Overlap / SampleRate;
for (int i = 0; i < cutLogarithmizedSpectrum.Count; i++)
{
Assert.IsTrue(System.Math.Abs(cutLogarithmizedSpectrum[i].StartsAt - (i * lengthOfOneFingerprint / 2)) < Epsilon);
}
}