/// <summary>
/// Method to analyze and add using the soundfingerprinting methods
/// </summary>
/// <param name="filePath">full file path</param>
/// <param name="repository">Soundfingerprinting Repository</param>
/// <param name="doOutputDebugInfo">decide whether to output debug info like spectrogram and audiofile (default value can be set)</param>
/// <param name="useHaarWavelet">decide whether to use haar wavelet compression or DCT compression</param>
/// <returns>true if successful</returns>
public static bool AnalyzeAndAddSoundfingerprinting(FileInfo filePath, Repository repository, bool doOutputDebugInfo=DEFAULT_DEBUG_INFO, bool useHaarWavelet = true)
{
DbgTimer t = new DbgTimer();
t.Start ();
// get work config from the audio file
WorkUnitParameterObject param = GetWorkUnitParameterObjectFromAudioFile(filePath);
param.FingerprintingConfiguration = fingerprintingConfigCreation;
string fileName = param.FileName;
// build track
Track track = new Track();
track.Title = param.FileName;
track.TrackLengthMs = (int) param.DurationInMs;
track.FilePath = param.PathToAudioFile;
track.Tags = param.Tags;
track.Id = -1; // this will be set by the insert method
// Get fingerprint signatures using the Soundfingerprinting methods
double[][] logSpectrogram;
List<bool[]> fingerprints;
List<double[][]> spectralImages;
if (repository.InsertTrackInDatabaseUsingSamples(track, param.FingerprintingConfiguration.NumberOfHashTables, param.FingerprintingConfiguration.NumberOfKeys, param, out logSpectrogram, out fingerprints, out spectralImages)) {
// store logSpectrogram as Matrix
Comirva.Audio.Util.Maths.Matrix logSpectrogramMatrix = new Comirva.Audio.Util.Maths.Matrix(logSpectrogram);
logSpectrogramMatrix = logSpectrogramMatrix.Transpose();
#region Debug for Soundfingerprinting Method
if (doOutputDebugInfo) {
// Image Service
ImageService imageService = new ImageService(repository.FingerprintService.SpectrumService, repository.FingerprintService.WaveletService);
imageService.GetLogSpectralImages(logSpectrogram, fingerprintingConfigCreation.Stride, fingerprintingConfigCreation.FingerprintLength, fingerprintingConfigCreation.Overlap, 2).Save(fileName + "_specgram_logimages.png");
logSpectrogramMatrix.DrawMatrixImageLogValues(fileName + "_specgram_logimage.png", true);
if (DEBUG_OUTPUT_TEXT) {
logSpectrogramMatrix.WriteCSV(fileName + "_specgram_log.csv", ";");
}
}
#endregion
} else {
// failed
Console.Out.WriteLine("Failed! Could not compute the soundfingerprint {0}!", fileName);
return false;
}
Dbg.WriteLine ("AnalyzeAndAddSoundfingerprinting - Total Execution Time: {0} ms", t.Stop().TotalMilliseconds);
return true;
}
//private static Mfcc mfccOptimized = new Mfcc(WINDOW_SIZE, SAMPLING_RATE, MEL_COEFFICIENTS, MFCC_COEFFICIENTS);
//private static MFCC mfccComirva = new MFCC(SAMPLING_RATE, WINDOW_SIZE, MFCC_COEFFICIENTS, true, 20.0, SAMPLING_RATE/2, MEL_COEFFICIENTS);
#endif
#region Methods
public static bool AnalyzeAndAdd(FileInfo filePath, Db db, DatabaseService databaseService, bool doOutputDebugInfo=DEFAULT_DEBUG_INFO, bool useHaarWavelet = true)
{
DbgTimer t = new DbgTimer();
t.Start ();
float[] audiodata = AudioFileReader.Decode(filePath.FullName, SAMPLING_RATE, SECONDS_TO_ANALYZE);
if (audiodata == null || audiodata.Length == 0) {
Dbg.WriteLine("Error! - No Audio Found");
return false;
}
// Read TAGs using BASS
FindSimilar.AudioProxies.BassProxy bass = FindSimilar.AudioProxies.BassProxy.Instance;
Un4seen.Bass.AddOn.Tags.TAG_INFO tag_info = bass.GetTagInfoFromFile(filePath.FullName);
// Name of file being processed
string name = StringUtils.RemoveNonAsciiCharacters(Path.GetFileNameWithoutExtension(filePath.Name));
#if DEBUG
if (Analyzer.DEBUG_INFO_VERBOSE) {
if (DEBUG_OUTPUT_TEXT) WriteAscii(audiodata, name + "_audiodata.ascii");
if (DEBUG_OUTPUT_TEXT) WriteF3Formatted(audiodata, name + "_audiodata.txt");
}
#endif
if (doOutputDebugInfo) {
DrawGraph(MathUtils.FloatToDouble(audiodata), name + "_audiodata.png");
}
// Calculate duration in ms
double duration = (double) audiodata.Length / SAMPLING_RATE * 1000;
// Explode samples to the range of 16 bit shorts (–32,768 to 32,767)
// Matlab multiplies with 2^15 (32768)
// e.g. if( max(abs(speech))<=1 ), speech = speech * 2^15; end;
MathUtils.Multiply(ref audiodata, AUDIO_MULTIPLIER); // 65536
// zero pad if the audio file is too short to perform a mfcc
if (audiodata.Length < (fingerprintingConfig.WdftSize + fingerprintingConfig.Overlap))
{
int lenNew = fingerprintingConfig.WdftSize + fingerprintingConfig.Overlap;
Array.Resize<float>(ref audiodata, lenNew);
}
// Get fingerprint signatures using the Soundfingerprinting methods
IPermutations permutations = new LocalPermutations("Soundfingerprinting\\perms.csv", ",");
Repository repository = new Repository(permutations, databaseService, fingerprintService);
// Image Service
ImageService imageService = new ImageService(
fingerprintService.SpectrumService,
fingerprintService.WaveletService);
// work config
WorkUnitParameterObject param = new WorkUnitParameterObject();
param.FingerprintingConfiguration = fingerprintingConfig;
param.AudioSamples = audiodata;
param.PathToAudioFile = filePath.FullName;
param.MillisecondsToProcess = SECONDS_TO_ANALYZE * 1000;
param.StartAtMilliseconds = 0;
// build track
Track track = new Track();
track.Title = name;
track.TrackLengthMs = (int) duration;
track.FilePath = filePath.FullName;
track.Id = -1; // this will be set by the insert method
#region parse tag_info
if (tag_info != null) {
Dictionary<string, string> tags = new Dictionary<string, string>();
//if (tag_info.title != string.Empty) tags.Add("title", tag_info.title);
if (tag_info.artist != string.Empty) tags.Add("artist", tag_info.artist);
if (tag_info.album != string.Empty) tags.Add("album", tag_info.album);
if (tag_info.albumartist != string.Empty) tags.Add("albumartist", tag_info.albumartist);
if (tag_info.year != string.Empty) tags.Add("year", tag_info.year);
if (tag_info.comment != string.Empty) tags.Add("comment", tag_info.comment);
if (tag_info.genre != string.Empty) tags.Add("genre", tag_info.genre);
if (tag_info.track != string.Empty) tags.Add("track", tag_info.track);
if (tag_info.disc != string.Empty) tags.Add("disc", tag_info.disc);
if (tag_info.copyright != string.Empty) tags.Add("copyright", tag_info.copyright);
if (tag_info.encodedby != string.Empty) tags.Add("encodedby", tag_info.encodedby);
if (tag_info.composer != string.Empty) tags.Add("composer", tag_info.composer);
if (tag_info.publisher != string.Empty) tags.Add("publisher", tag_info.publisher);
if (tag_info.lyricist != string.Empty) tags.Add("lyricist", tag_info.lyricist);
if (tag_info.remixer != string.Empty) tags.Add("remixer", tag_info.remixer);
if (tag_info.producer != string.Empty) tags.Add("producer", tag_info.producer);
if (tag_info.bpm != string.Empty) tags.Add("bpm", tag_info.bpm);
//if (tag_info.filename != string.Empty) tags.Add("filename", tag_info.filename);
tags.Add("channelinfo", tag_info.channelinfo.ToString());
//if (tag_info.duration > 0) tags.Add("duration", tag_info.duration.ToString());
if (tag_info.bitrate > 0) tags.Add("bitrate", tag_info.bitrate.ToString());
if (tag_info.replaygain_track_gain != -100f) tags.Add("replaygain_track_gain", tag_info.replaygain_track_gain.ToString());
if (tag_info.replaygain_track_peak != -1f) tags.Add("replaygain_track_peak", tag_info.replaygain_track_peak.ToString());
if (tag_info.conductor != string.Empty) tags.Add("conductor", tag_info.conductor);
if (tag_info.grouping != string.Empty) tags.Add("grouping", tag_info.grouping);
if (tag_info.mood != string.Empty) tags.Add("mood", tag_info.mood);
if (tag_info.rating != string.Empty) tags.Add("rating", tag_info.rating);
if (tag_info.isrc != string.Empty) tags.Add("isrc", tag_info.isrc);
foreach(var nativeTag in tag_info.NativeTags) {
string[] keyvalue = nativeTag.Split('=');
tags.Add(keyvalue[0], keyvalue[1]);
}
track.Tags = tags;
}
#endregion
double[][] logSpectrogram;
if (repository.InsertTrackInDatabaseUsingSamples(track, 25, 4, param, out logSpectrogram)) {
// store logSpectrogram as Matrix
Comirva.Audio.Util.Maths.Matrix logSpectrogramMatrix = new Comirva.Audio.Util.Maths.Matrix(logSpectrogram);
logSpectrogramMatrix = logSpectrogramMatrix.Transpose();
#region Debug for Soundfingerprinting Method
if (doOutputDebugInfo) {
imageService.GetLogSpectralImages(logSpectrogram, fingerprintingConfig.Stride, fingerprintingConfig.FingerprintLength, fingerprintingConfig.Overlap, 2).Save(name + "_specgram_logimages.png");
logSpectrogramMatrix.DrawMatrixImageLogValues(name + "_specgram_logimage.png", true);
if (DEBUG_OUTPUT_TEXT) {
logSpectrogramMatrix.WriteCSV(name + "_specgram_log.csv", ";");
}
}
#endregion
#region Insert Statistical Cluster Model Similarity Audio Feature as well
Comirva.Audio.Util.Maths.Matrix scmsMatrix = null;
if (useHaarWavelet) {
#region Wavelet Transform
int lastHeight = 0;
int lastWidth = 0;
scmsMatrix = mfccMirage.ApplyWaveletCompression(ref logSpectrogramMatrix, out lastHeight, out lastWidth);
#if DEBUG
if (Analyzer.DEBUG_INFO_VERBOSE) {
if (DEBUG_OUTPUT_TEXT) scmsMatrix.WriteAscii(name + "_waveletdata.ascii");
}
#endif
if (doOutputDebugInfo) {
scmsMatrix.DrawMatrixImageLogValues(name + "_waveletdata.png", true);
}
#if DEBUG
if (Analyzer.DEBUG_INFO_VERBOSE) {
#region Inverse Wavelet
// try to do an inverse wavelet transform
Comirva.Audio.Util.Maths.Matrix stftdata_inverse_wavelet = mfccMirage.InverseWaveletCompression(ref scmsMatrix, lastHeight, lastWidth, logSpectrogramMatrix.Rows, logSpectrogramMatrix.Columns);
if (DEBUG_OUTPUT_TEXT) stftdata_inverse_wavelet.WriteCSV(name + "_specgramlog_inverse_wavelet.csv", ";");
stftdata_inverse_wavelet.DrawMatrixImageLogValues(name + "_specgramlog_inverse_wavelet.png", true);
#endregion
}
#endif
#endregion
} else {
#region DCT Transform
// It seems the Mirage way of applying the DCT is slightly faster than the
// Comirva way due to less loops
scmsMatrix = mfccMirage.ApplyDCT(ref logSpectrogramMatrix);
#if DEBUG
if (Analyzer.DEBUG_INFO_VERBOSE) {
if (DEBUG_OUTPUT_TEXT) scmsMatrix.WriteAscii(name + "_mfccdata.ascii");
}
#endif
if (doOutputDebugInfo) {
scmsMatrix.DrawMatrixImageLogValues(name + "_mfccdata.png", true);
}
#if DEBUG
if (Analyzer.DEBUG_INFO_VERBOSE) {
#region Inverse MFCC
// try to do an inverse mfcc
Comirva.Audio.Util.Maths.Matrix stftdata_inverse_mfcc = mfccMirage.InverseDCT(ref scmsMatrix);
if (DEBUG_OUTPUT_TEXT) stftdata_inverse_mfcc.WriteCSV(name + "_stftdata_inverse_mfcc.csv", ";");
stftdata_inverse_mfcc.DrawMatrixImageLogValues(name + "_specgramlog_inverse_mfcc.png", true);
#endregion
}
#endif
#endregion
}
// Store in a Statistical Cluster Model Similarity class.
// A Gaussian representation of a song
Scms audioFeature = Scms.GetScms(scmsMatrix, name);
if (audioFeature != null) {
// Store image if debugging
if (doOutputDebugInfo) {
audioFeature.Image = scmsMatrix.DrawMatrixImageLogValues(name + "_featuredata.png", true, false, 0, 0, true);
}
// Store bitstring hash as well
string hashString = GetBitString(scmsMatrix);
audioFeature.BitString = hashString;
// Store duration
audioFeature.Duration = (long) duration;
// Store file name
audioFeature.Name = filePath.FullName;
int id = track.Id;
if (db.AddTrack(ref id, audioFeature) == -1) {
Console.Out.WriteLine("Failed! Could not add audioFeature to database {0}!", name);
}
}
#endregion
} else {
// failed
return false;
}
Dbg.WriteLine ("AnalyzeAndAdd - Total Execution Time: {0} ms", t.Stop().TotalMilliseconds);
return true;
}
public static AudioFeature AnalyzeSoundfingerprinting(FileInfo filePath, bool doOutputDebugInfo=DEFAULT_DEBUG_INFO, bool useHaarWavelet = true)
{
DbgTimer t = new DbgTimer();
t.Start ();
float[] audiodata = AudioFileReader.Decode(filePath.FullName, SAMPLING_RATE, SECONDS_TO_ANALYZE);
if (audiodata == null || audiodata.Length == 0) {
Dbg.WriteLine("Error! - No Audio Found");
return null;
}
// Read TAGs using BASS
FindSimilar.AudioProxies.BassProxy bass = FindSimilar.AudioProxies.BassProxy.Instance;
Un4seen.Bass.AddOn.Tags.TAG_INFO tag_info = bass.GetTagInfoFromFile(filePath.FullName);
// Name of file being processed
string name = StringUtils.RemoveNonAsciiCharacters(Path.GetFileNameWithoutExtension(filePath.Name));
#if DEBUG
if (Analyzer.DEBUG_INFO_VERBOSE) {
if (DEBUG_OUTPUT_TEXT) WriteAscii(audiodata, name + "_audiodata.ascii");
if (DEBUG_OUTPUT_TEXT) WriteF3Formatted(audiodata, name + "_audiodata.txt");
}
#endif
if (doOutputDebugInfo) {
DrawGraph(MathUtils.FloatToDouble(audiodata), name + "_audiodata.png");
}
// Calculate duration in ms
double duration = (double) audiodata.Length / SAMPLING_RATE * 1000;
// zero pad if the audio file is too short to perform a mfcc
if (audiodata.Length < (fingerprintingConfig.WdftSize + fingerprintingConfig.Overlap))
{
int lenNew = fingerprintingConfig.WdftSize + fingerprintingConfig.Overlap;
Array.Resize<float>(ref audiodata, lenNew);
}
// Get fingerprint signatures using the Soundfingerprinting methods
// Get database
DatabaseService databaseService = DatabaseService.Instance;
IPermutations permutations = new LocalPermutations("Soundfingerprinting\\perms.csv", ",");
Repository repository = new Repository(permutations, databaseService, fingerprintService);
// Image Service
ImageService imageService = new ImageService(
fingerprintService.SpectrumService,
fingerprintService.WaveletService);
// work config
WorkUnitParameterObject param = new WorkUnitParameterObject();
param.FingerprintingConfiguration = fingerprintingConfig;
param.AudioSamples = audiodata;
param.PathToAudioFile = filePath.FullName;
param.MillisecondsToProcess = SECONDS_TO_ANALYZE * 1000;
param.StartAtMilliseconds = 0;
// build track
Track track = new Track();
track.Title = name;
track.TrackLengthMs = (int) duration;
track.FilePath = filePath.FullName;
track.Id = -1; // this will be set by the insert method
#region parse tag_info
if (tag_info != null) {
Dictionary<string, string> tags = new Dictionary<string, string>();
//if (tag_info.title != string.Empty) tags.Add("title", tag_info.title);
if (tag_info.artist != string.Empty) tags.Add("artist", tag_info.artist);
if (tag_info.album != string.Empty) tags.Add("album", tag_info.album);
if (tag_info.albumartist != string.Empty) tags.Add("albumartist", tag_info.albumartist);
if (tag_info.year != string.Empty) tags.Add("year", tag_info.year);
if (tag_info.comment != string.Empty) tags.Add("comment", tag_info.comment);
if (tag_info.genre != string.Empty) tags.Add("genre", tag_info.genre);
if (tag_info.track != string.Empty) tags.Add("track", tag_info.track);
if (tag_info.disc != string.Empty) tags.Add("disc", tag_info.disc);
if (tag_info.copyright != string.Empty) tags.Add("copyright", tag_info.copyright);
if (tag_info.encodedby != string.Empty) tags.Add("encodedby", tag_info.encodedby);
if (tag_info.composer != string.Empty) tags.Add("composer", tag_info.composer);
if (tag_info.publisher != string.Empty) tags.Add("publisher", tag_info.publisher);
if (tag_info.lyricist != string.Empty) tags.Add("lyricist", tag_info.lyricist);
if (tag_info.remixer != string.Empty) tags.Add("remixer", tag_info.remixer);
if (tag_info.producer != string.Empty) tags.Add("producer", tag_info.producer);
if (tag_info.bpm != string.Empty) tags.Add("bpm", tag_info.bpm);
//if (tag_info.filename != string.Empty) tags.Add("filename", tag_info.filename);
tags.Add("channelinfo", tag_info.channelinfo.ToString());
//if (tag_info.duration > 0) tags.Add("duration", tag_info.duration.ToString());
if (tag_info.bitrate > 0) tags.Add("bitrate", tag_info.bitrate.ToString());
if (tag_info.replaygain_track_gain != -100f) tags.Add("replaygain_track_gain", tag_info.replaygain_track_gain.ToString());
if (tag_info.replaygain_track_peak != -1f) tags.Add("replaygain_track_peak", tag_info.replaygain_track_peak.ToString());
if (tag_info.conductor != string.Empty) tags.Add("conductor", tag_info.conductor);
if (tag_info.grouping != string.Empty) tags.Add("grouping", tag_info.grouping);
if (tag_info.mood != string.Empty) tags.Add("mood", tag_info.mood);
if (tag_info.rating != string.Empty) tags.Add("rating", tag_info.rating);
if (tag_info.isrc != string.Empty) tags.Add("isrc", tag_info.isrc);
foreach(var nativeTag in tag_info.NativeTags) {
string[] keyvalue = nativeTag.Split('=');
tags.Add(keyvalue[0], keyvalue[1]);
}
track.Tags = tags;
}
#endregion
AudioFeature audioFeature = null;
double[][] logSpectrogram;
if (repository.InsertTrackInDatabaseUsingSamples(track, 25, 4, param, out logSpectrogram)) {
if (doOutputDebugInfo) {
imageService.GetLogSpectralImages(logSpectrogram, fingerprintingConfig.Stride, fingerprintingConfig.FingerprintLength, fingerprintingConfig.Overlap, 2).Save(name + "_specgram_logimages.png");
Comirva.Audio.Util.Maths.Matrix logSpectrogramMatrix = new Comirva.Audio.Util.Maths.Matrix(logSpectrogram);
logSpectrogramMatrix = logSpectrogramMatrix.Transpose();
logSpectrogramMatrix.DrawMatrixImageLogValues(name + "_specgram_logimage.png", true);
if (DEBUG_OUTPUT_TEXT) {
logSpectrogramMatrix.WriteCSV(name + "_specgram_log.csv", ";");
}
}
audioFeature = new DummyAudioFeature();
// Store duration
audioFeature.Duration = (long) duration;
// Store file name
audioFeature.Name = filePath.FullName;
} else {
// failed
}
Dbg.WriteLine ("Soundfingerprinting - Total Execution Time: {0} ms", t.Stop().TotalMilliseconds);
return audioFeature;
}
/// <summary>
/// Method to analyse and add all the different types of audio features
/// </summary>
/// <param name="filePath">full file path</param>
/// <param name="db">Scms database (Mirage)</param>
/// <param name="repository">Soundfingerprinting Repository</param>
/// <param name="doOutputDebugInfo">decide whether to output debug info like spectrogram and audiofile (default value can be set)</param>
/// <param name="useHaarWavelet">decide whether to use haar wavelet compression or DCT compression</param>
/// <returns>true if successful</returns>
public static bool AnalyzeAndAddComplete(FileInfo filePath, Db db, Repository repository, bool doOutputDebugInfo=DEFAULT_DEBUG_INFO, bool useHaarWavelet = true) {
DbgTimer t = new DbgTimer();
t.Start ();
// get work config from the audio file
WorkUnitParameterObject param = GetWorkUnitParameterObjectFromAudioFile(filePath);
if (param == null) return false;
param.FingerprintingConfiguration = fingerprintingConfigCreation;
string fileName = param.FileName;
// build track
Track track = new Track();
track.Title = param.FileName;
track.TrackLengthMs = (int) param.DurationInMs;
track.FilePath = param.PathToAudioFile;
track.Tags = param.Tags;
track.Id = -1; // this will be set by the insert method
double[][] logSpectrogram;
List<bool[]> fingerprints;
if (repository.InsertTrackInDatabaseUsingSamples(track, param.FingerprintingConfiguration.NumberOfHashTables, param.FingerprintingConfiguration.NumberOfKeys, param, out logSpectrogram, out fingerprints)) {
// store logSpectrogram as Matrix
try {
Comirva.Audio.Util.Maths.Matrix logSpectrogramMatrix = new Comirva.Audio.Util.Maths.Matrix(logSpectrogram);
logSpectrogramMatrix = logSpectrogramMatrix.Transpose();
#region Output debugging information (Saving spectrograms and/or csv files)
if (doOutputDebugInfo) {
logSpectrogramMatrix.DrawMatrixImageLogValues(fileName + "_matrix_spectrogram.png", true);
if (DEBUG_OUTPUT_TEXT) {
logSpectrogramMatrix.WriteCSV(fileName + "_matrix_spectrogram.csv", ";");
}
// Save debug images using fingerprinting methods
SaveFingerprintingDebugImages(fileName, logSpectrogram, fingerprints, repository.FingerprintService, param.FingerprintingConfiguration);
}
#endregion
// Insert Statistical Cluster Model Similarity Audio Feature as well
if (!AnalyseAndAddScmsUsingLogSpectrogram(logSpectrogramMatrix, param, db, track.Id, doOutputDebugInfo, useHaarWavelet)) {
Dbg.WriteLine("AnalyzeAndAddComplete - Failed inserting Statistical Cluster Model Similarity Audio Feature");
// Failed, but ignore!
}
} catch (Exception e) {
Dbg.WriteLine("AnalyzeAndAddComplete - Failed creating Statistical Cluster Model Similarity Audio Feature");
Dbg.WriteLine(e.Message);
// Failed, but ignore!
}
} else {
// Failed
return false;
}
Dbg.WriteLine("AnalyzeAndAddComplete - Total Execution Time: {0} ms", t.Stop().TotalMilliseconds);
return true;
}