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 void SubFingerprintsAreBuiltCorrectlyFromAudioSamplesForTrack()
{
const int TenSeconds = 10;
var samples = new AudioSamples { Samples = TestUtilities.GenerateRandomFloatArray(SampleRate * TenSeconds) };
const int ThreeFingerprints = 3;
var rawFingerprints = GetGenericFingerprints(ThreeFingerprints);
fingerprintService.Setup(service => service.CreateFingerprints(samples, It.IsAny<DefaultFingerprintConfiguration>())).Returns(rawFingerprints);
lshAlgorithm.Setup(service => service.Hash(It.IsAny<Fingerprint>(), NumberOfHashTables, NumberOfHashKeysPerTable, It.IsAny<IEnumerable<string>>())).Returns(new HashedFingerprint(GenericSignature(), GenericHashBuckets(), 0, 0, Enumerable.Empty<string>()));
var hashDatas = fingerprintCommandBuilder.BuildFingerprintCommand()
.From(samples)
.UsingServices(audioService.Object)
.Hash()
.Result;
Assert.AreEqual(ThreeFingerprints, hashDatas.Count);
byte[] genericSignature = GenericSignature();
long[] genericHashBuckets = GenericHashBuckets();
foreach (var hashData in hashDatas)
{
CollectionAssert.AreEqual(genericSignature, hashData.SubFingerprint);
CollectionAssert.AreEqual(genericHashBuckets, hashData.HashBins);
}
}
public void SubFingerprintsAreBuiltCorrectlyFromFileForTrack()
{
const int TenSeconds = 10;
var samples = new AudioSamples { Samples = TestUtilities.GenerateRandomFloatArray(SampleRate * TenSeconds) };
const int ThreeFingerprints = 3;
var rawFingerprints = GetGenericFingerprints(ThreeFingerprints);
const string PathToAudioFile = "path-to-audio-file";
audioService.Setup(service => service.ReadMonoSamplesFromFile(PathToAudioFile, SampleRate)).Returns(samples);
fingerprintService.Setup(
service => service.CreateFingerprints(samples, It.IsAny<DefaultFingerprintConfiguration>())).Returns(
rawFingerprints);
lshAlgorithm.Setup(service => service.Hash(It.IsAny<Fingerprint>(), NumberOfHashTables, NumberOfHashKeysPerTable)).Returns(new HashedFingerprint(GenericSignature, GenericHashBuckets, 0, 0.928));
var hashDatas = fingerprintCommandBuilder.BuildFingerprintCommand()
.From(PathToAudioFile)
.UsingServices(audioService.Object)
.Hash()
.Result;
Assert.AreEqual(ThreeFingerprints, hashDatas.Count);
foreach (var hashData in hashDatas)
{
Assert.AreSame(GenericSignature, hashData.SubFingerprint);
}
}
private AudioSamples Copy(AudioSamples chunk)
{
float[] withCached = new float[tailLength + chunk.Samples.Length];
Buffer.BlockCopy(tailBuffer, 0, withCached, 0, sizeof(float) * tailLength);
Buffer.BlockCopy(chunk.Samples, 0, withCached, sizeof(float) * tailLength, sizeof(float) * chunk.Samples.Length);
return(new AudioSamples(withCached, chunk.Origin, chunk.SampleRate, relativeTo));
}
private void NormalizeAudioIfNecessary(AudioSamples samples, FingerprintConfiguration configuration)
{
if (configuration.NormalizeSignal)
{
audioSamplesNormalizer.NormalizeInPlace(samples.Samples);
}
}
private void Cache(AudioSamples chunk)
{
int nextStride = Stride.NextStride;
tailLength = tailBuffer.Length - nextStride;
relativeTo = chunk.RelativeTo.AddSeconds((double)(chunk.Samples.Length - tailLength) / chunk.SampleRate);
Buffer.BlockCopy(chunk.Samples, sizeof(float) * (chunk.Samples.Length - tailLength), tailBuffer, 0, sizeof(float) * tailLength);
}
private void Cache(AudioSamples chunk)
{
int nextStride = Stride.NextStride;
left = chunk.Samples.Length - nextStride;
relativeTo = chunk.RelativeTo.AddSeconds((double)(nextStride) / chunk.SampleRate);
Buffer.BlockCopy(chunk.Samples, sizeof(float) * (nextStride), buffer, 0, sizeof(float) * left);
}
public List<SpectralImage> CreateLogSpectrogram(AudioSamples audioSamples, SpectrogramConfig configuration)
{
int width = (audioSamples.Samples.Length - configuration.WdftSize) / configuration.Overlap;
if (width < 1)
{
return new List<SpectralImage>();
}
float[][] frames = new float[width][];
int[] logFrequenciesIndexes = logUtility.GenerateLogFrequenciesRanges(audioSamples.SampleRate, configuration);
float[] window = configuration.Window.GetWindow(configuration.WdftSize);
Parallel.For(0, width, i => {
float[] complexSignal = fftService.FFTForward(audioSamples.Samples, i * configuration.Overlap, configuration.WdftSize, window);
frames[i] = ExtractLogBins(complexSignal, logFrequenciesIndexes, configuration.LogBins, configuration.WdftSize);
});
return CutLogarithmizedSpectrum(frames, audioSamples.SampleRate, configuration);
}
public void FingerprintsAreBuiltCorrectlyFromFile()
{
const string PathToAudioFile = "path-to-audio-file";
const int TenSeconds = 10;
var samples = new AudioSamples { Samples = TestUtilities.GenerateRandomFloatArray(SampleRate * TenSeconds) };
const int ThreeFingerprints = 3;
var rawFingerprints = GetGenericFingerprints(ThreeFingerprints);
audioService.Setup(service => service.ReadMonoSamplesFromFile(PathToAudioFile, SampleRate)).Returns(samples);
fingerprintService.Setup(service => service.CreateFingerprints(samples, It.IsAny<DefaultFingerprintConfiguration>())).Returns(rawFingerprints);
var fingerprints = ((FingerprintCommand)fingerprintCommandBuilder.BuildFingerprintCommand()
.From(PathToAudioFile)
.UsingServices(audioService.Object))
.Fingerprint()
.Result;
Assert.AreEqual(ThreeFingerprints, fingerprints.Count);
Assert.AreSame(rawFingerprints, fingerprints);
}
public AudioSamples Aggregate(AudioSamples chunk)
{
if (chunk.Samples.Length < MinSize)
{
throw new ArgumentException($"{nameof(chunk)} cannot be less than {MinSize}");
}
inc++;
if (inc == 0)
{
Cache(chunk);
return(chunk);
}
var cached = Copy(chunk);
Cache(chunk);
return(cached);
}
public void SubFingerprintsAreBuiltCorrectlyFromFileForTrackStartingAtSpecificSecond()
{
const string PathToAudioFile = "path-to-audio-file";
const int StartSecond = 10;
const int SecondsToProcess = 20;
AudioSamples samples = new AudioSamples
{ Samples = TestUtilities.GenerateRandomFloatArray(SampleRate * 10) };
const int ThreeFingerprints = 3;
var rawFingerprints = GetGenericFingerprints(ThreeFingerprints);
audioService.Setup(service => service.ReadMonoSamplesFromFile(PathToAudioFile, SampleRate, SecondsToProcess, StartSecond)).Returns(samples);
fingerprintService.Setup(service => service.CreateFingerprints(samples, It.IsAny<DefaultFingerprintConfiguration>())).Returns(rawFingerprints);
byte[] genericSignature = GenericSignature();
lshAlgorithm.Setup(service => service.Hash(It.IsAny<Fingerprint>(), NumberOfHashTables, NumberOfHashKeysPerTable, It.IsAny<IEnumerable<string>>())).Returns(
new HashedFingerprint(genericSignature, GenericHashBuckets(), 0, 0, Enumerable.Empty<string>()));
var hashDatas = fingerprintCommandBuilder.BuildFingerprintCommand()
.From(PathToAudioFile, SecondsToProcess, StartSecond)
.UsingServices(audioService.Object)
.Hash()
.Result;
Assert.AreEqual(ThreeFingerprints, hashDatas.Count);
foreach (var hashData in hashDatas)
{
CollectionAssert.AreEqual(genericSignature, hashData.SubFingerprint);
}
audioService.Verify(service => service.ReadMonoSamplesFromFile(PathToAudioFile, SampleRate, SecondsToProcess, StartSecond));
}
public List<Fingerprint> CreateFingerprints(AudioSamples samples, FingerprintConfiguration configuration)
{
NormalizeAudioIfNecessary(samples, configuration);
var spectrum = spectrumService.CreateLogSpectrogram(samples, configuration.SpectrogramConfig);
return CreateFingerprintsFromLogSpectrum(spectrum, configuration);
}
public IWithFingerprintConfiguration From(AudioSamples audioSamples)
{
createFingerprintsMethod = () => fingerprintService.CreateFingerprints(audioSamples, FingerprintConfiguration);
return this;
}
public IWithQueryAndFingerprintConfiguration From(AudioSamples audioSamples)
{
fingerprintingMethodFromSelector = () => fingerprintCommandBuilder.BuildFingerprintCommand().From(audioSamples);
return this;
}
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));
}
