public static List<FindSimilar.QueryResult> SimilarTracksSoundfingerprintingList(FileInfo filePath, Repository repository)
{
DbgTimer t = new DbgTimer();
t.Start ();
// get work config from the audio file
WorkUnitParameterObject param = GetWorkUnitParameterObjectFromAudioFile(filePath);
param.FingerprintingConfiguration = fingerprintingConfigQuerying;
// TODO: i don't really know how the threshold tables work.
// 1 returns more similar hits
// 2 returns sometimes only the one we search for
// even 0 seem to work (like 1)
List<FindSimilar.QueryResult> candidates = repository.FindSimilarFromAudioSamplesList(param.FingerprintingConfiguration.NumberOfHashTables, param.FingerprintingConfiguration.NumberOfKeys, 0, param);
Dbg.WriteLine ("SimilarTracksSoundfingerprintingList - Total Execution Time: {0} ms", t.Stop().TotalMilliseconds);
return candidates;
}
public FindSimilarClientForm()
{
//
// The InitializeComponent() call is required for Windows Forms designer support.
//
InitializeComponent();
//
// Constructor code after the InitializeComponent() call.
//
this.version.Text = Mirage.Mir.VERSION;
this.DistanceTypeCombo.DataSource = Enum.GetValues(typeof(AudioFeature.DistanceType));
/*
this.dataGridView1.Columns.Add("Id", "Id");
this.dataGridView1.Columns[0].AutoSizeMode = DataGridViewAutoSizeColumnMode.AllCells;
this.dataGridView1.Columns.Add("Path", "Path");
this.dataGridView1.Columns[1].AutoSizeMode = DataGridViewAutoSizeColumnMode.Fill;
this.dataGridView1.Columns.Add("Duration_Similarity", "Duration (ms) / Similarity");
this.dataGridView1.Columns[2].AutoSizeMode = DataGridViewAutoSizeColumnMode.AllCells;
*/
this.db = new Db();
// Instansiate soundfingerprinting Repository
FingerprintService fingerprintService = Analyzer.GetSoundfingerprintingService();
this.databaseService = DatabaseService.Instance;
//IPermutations permutations = new LocalPermutations("Soundfingerprinting\\perms.csv", ",");
IPermutations permutations = new LocalPermutations("perms-new.csv", ",");
repository = new Repository(permutations, databaseService, fingerprintService);
if (rbScms.Checked) {
IgnoreFileLengthCheckBox.Visible = true;
DistanceTypeCombo.Visible = true;
} else {
IgnoreFileLengthCheckBox.Visible = false;
DistanceTypeCombo.Visible = false;
}
ReadAllTracks();
}
/// <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;
}
/// <summary>
/// Query the database for perceptually similar tracks using the sound fingerprinting methods
/// </summary>
/// <param name="filePath">input file</param>
/// <returns>a dictionary of similar tracks</returns>
public static Dictionary<Track, double> SimilarTracksSoundfingerprinting(FileInfo filePath, Repository repository)
{
DbgTimer t = new DbgTimer();
t.Start ();
// get work config from the audio file
WorkUnitParameterObject param = GetWorkUnitParameterObjectFromAudioFile(filePath);
param.FingerprintingConfiguration = fingerprintingConfigQuerying;
Dictionary<Track, double> candidates = repository.FindSimilarFromAudioSamples(param.FingerprintingConfiguration.NumberOfHashTables, param.FingerprintingConfiguration.NumberOfKeys, 1, param);
Dbg.WriteLine ("SimilarTracksSoundfingerprinting - Total Execution Time: {0} ms", t.Stop().TotalMilliseconds);
return candidates;
}
/// <summary>
/// Scan a directory recursively and add all the audio files found to the database
/// </summary>
/// <param name="path">Path to directory</param>
/// <param name="db">MandelEllis or Scms Database Instance</param>
/// <param name="repository">Soundfingerprinting Repository</param>
/// <param name="skipDurationAboveSeconds">Skip files with duration longer than this number of seconds (0 or less disables this)</param>
/// <param name="silent">true if silent mode (reduced console output)</param>
public static void ScanDirectory(string path, Db db, Repository repository, double skipDurationAboveSeconds, bool silent=false) {
Stopwatch stopWatch = Stopwatch.StartNew();
FAILED_FILES_LOG.Delete();
WARNING_FILES_LOG.Delete();
// scan directory for audio files
try
{
// By some reason the IOUtils.GetFiles returns a higher count than what seams correct?!
IEnumerable<string> filesAll =
Directory.EnumerateFiles(path, "*", SearchOption.AllDirectories)
.Where(f => extensions.Contains(Path.GetExtension(f).ToLower()));
Console.Out.WriteLine("Found {0} files in scan directory.", filesAll.Count());
// Get all already processed files stored in the database and store in memory
// It seems to work well with huge volumes of files (200k)
IList<string> filesAlreadyProcessed = repository.DatabaseService.ReadTrackFilenames();
Console.Out.WriteLine("Database contains {0} already processed files.", filesAlreadyProcessed.Count);
// find the files that has not already been added to the database
List<string> filesRemaining = filesAll.Except(filesAlreadyProcessed).ToList();
Console.Out.WriteLine("Found {0} files remaining in scan directory to be processed.", filesRemaining.Count);
int filesCounter = 1;
int filesAllCounter = filesAlreadyProcessed.Count + 1;
#if !DEBUG
Console.Out.WriteLine("Running in multi-threaded mode!");
Parallel.ForEach(filesRemaining, file =>
{
#else
Console.Out.WriteLine("Running in single-threaded mode!");
foreach (string file in filesRemaining)
{
#endif
FileInfo fileInfo = new FileInfo(file);
// Try to use Un4Seen Bass to check duration
BassProxy bass = BassProxy.Instance;
double duration = bass.GetDurationInSeconds(fileInfo.FullName);
// check if we should skip files longer than x seconds
if ( (skipDurationAboveSeconds > 0 && duration > 0 && duration < skipDurationAboveSeconds)
|| skipDurationAboveSeconds <= 0
|| duration < 0) {
if(!Analyzer.AnalyzeAndAddComplete(fileInfo, db, repository)) {
//if(!Analyzer.AnalyzeAndAddSoundfingerprinting(fileInfo, repository)) {
//if(!Analyzer.AnalyzeAndAddScms(fileInfo, db)) {
Console.Out.WriteLine("Failed! Could not generate audio fingerprint for {0}!", fileInfo.Name);
IOUtils.LogMessageToFile(FAILED_FILES_LOG, fileInfo.FullName);
} else {
Console.Out.WriteLine("[{1}/{2} - {3}/{4}] Succesfully added {0} to database. (Thread: {5})", fileInfo.Name, filesCounter, filesRemaining.Count, filesAllCounter, filesAll.Count(), Thread.CurrentThread.ManagedThreadId);
// Threadsafe increment (TODO: doesn't always seem to work?)
//filesCounter++;
//filesAllCounter++;
Interlocked.Increment(ref filesCounter);
Interlocked.Increment(ref filesAllCounter);
}
} else {
if (!silent) Console.Out.WriteLine("Skipping {0} since duration exceeds limit ({1:0.00} > {2:0.00} sec.)", fileInfo.Name, duration, skipDurationAboveSeconds);
}
fileInfo = null;
}
#if !DEBUG
);
#endif
int filesActuallyProcessed = filesCounter -1;
Console.WriteLine("Added {0} out of a total remaining set of {1} files. (Of {2} files found).", filesActuallyProcessed, filesRemaining.Count(), filesAll.Count());
}
catch (UnauthorizedAccessException UAEx)
{
Console.WriteLine(UAEx.Message);
}
catch (PathTooLongException PathEx)
{
Console.WriteLine(PathEx.Message);
}
catch (System.NullReferenceException NullEx) {
Console.WriteLine(NullEx.Message);
}
Console.WriteLine("Time used: {0}", stopWatch.Elapsed);
}
public static void Main(string[] args) {
Analyzer.AnalysisMethod analysisMethod = Analyzer.AnalysisMethod.SCMS;
//Analyzer.AnalysisMethod analysisMethod = Analyzer.AnalysisMethod.MandelEllis;
string scanPath = "";
double skipDurationAboveSeconds = -1; // less than zero disables this
string queryPath = "";
int queryId = -1;
int numToTake = 20;
double percentage = 0.4; // percentage below and above when querying
bool resetdb = false;
bool silent = false;
AudioFeature.DistanceType distanceType = AudioFeature.DistanceType.KullbackLeiblerDivergence;
// Command line parsing
Arguments CommandLine = new Arguments(args);
if(CommandLine["match"] != null) {
queryPath = CommandLine["match"];
}
if(CommandLine["matchid"] != null) {
string matchId = CommandLine["matchid"];
queryId = int.Parse(matchId);
}
if(CommandLine["scandir"] != null) {
scanPath = CommandLine["scandir"];
}
if(CommandLine["skipduration"] != null) {
double.TryParse(CommandLine["skipduration"], NumberStyles.Number,CultureInfo.InvariantCulture, out skipDurationAboveSeconds);
}
if(CommandLine["num"] != null) {
string num = CommandLine["num"];
numToTake = int.Parse(num);
}
if(CommandLine["percentage"] != null) {
double.TryParse(CommandLine["percentage"], NumberStyles.Number,CultureInfo.InvariantCulture, out percentage);
}
if(CommandLine["type"] != null) {
string type = CommandLine["type"];
if (type.Equals("kl", StringComparison.InvariantCultureIgnoreCase)) {
distanceType = AudioFeature.DistanceType.KullbackLeiblerDivergence;
} else if (type.StartsWith("dtw", StringComparison.InvariantCultureIgnoreCase)) {
if (type.Equals("dtwe", StringComparison.InvariantCultureIgnoreCase)) {
distanceType = AudioFeature.DistanceType.Dtw_Euclidean;
} else if (type.Equals("dtwe2", StringComparison.InvariantCultureIgnoreCase)) {
distanceType = AudioFeature.DistanceType.Dtw_SquaredEuclidean;
} else if (type.Equals("dtwman", StringComparison.InvariantCultureIgnoreCase)) {
distanceType = AudioFeature.DistanceType.Dtw_Manhattan;
} else if (type.Equals("dtwmax", StringComparison.InvariantCultureIgnoreCase)) {
distanceType = AudioFeature.DistanceType.Dtw_Maximum;
} else if (type.Equals("ucrdtw", StringComparison.InvariantCultureIgnoreCase)) {
distanceType = AudioFeature.DistanceType.UCR_Dtw;
} else {
distanceType = AudioFeature.DistanceType.Dtw_Euclidean;
}
}
}
if(CommandLine["dtw"] != null || CommandLine["dtwe"] != null) {
distanceType = AudioFeature.DistanceType.Dtw_Euclidean;
}
if(CommandLine["dtwe2"] != null) {
distanceType = AudioFeature.DistanceType.Dtw_SquaredEuclidean;
}
if(CommandLine["dtwman"] != null) {
distanceType = AudioFeature.DistanceType.Dtw_Manhattan;
}
if(CommandLine["dtwmax"] != null) {
distanceType = AudioFeature.DistanceType.Dtw_Maximum;
}
if(CommandLine["kl"] != null) {
distanceType = AudioFeature.DistanceType.KullbackLeiblerDivergence;
}
if(CommandLine["ucrdtw"] != null) {
distanceType = AudioFeature.DistanceType.UCR_Dtw;
}
if(CommandLine["resetdb"] != null) {
resetdb = true;
}
if(CommandLine["silent"] != null) {
silent = true;
}
if(CommandLine["permutations"] != null) {
Console.WriteLine("Generating hash permutations for used by the Soundfingerprinting methods.");
Console.WriteLine("Saving to file: {0}", "Soundfingerprinting\\perms-new.csv");
Console.WriteLine();
PermutationGeneratorService permutationGeneratorService = new PermutationGeneratorService();
Analyzer.GenerateAndSavePermutations(permutationGeneratorService, "Soundfingerprinting\\perms-new.csv");
return;
}
if(CommandLine["?"] != null) {
PrintUsage();
return;
}
if(CommandLine["help"] != null) {
PrintUsage();
return;
}
if(CommandLine["gui"] != null) {
StartGUI();
return;
}
if (queryPath == "" && queryId == -1 && scanPath == "") {
PrintUsage();
return;
}
// Get database
Db mandelEllisScmsDatabase = new Db(); // For MandelEllis and SCMS
// Instansiate soundfingerprinting Repository
FingerprintService fingerprintService = Analyzer.GetSoundfingerprintingService();
DatabaseService databaseService = DatabaseService.Instance; // For AudioFingerprinting
IPermutations permutations = new LocalPermutations("Soundfingerprinting\\perms.csv", ",");
//IPermutations permutations = new LocalPermutations("Soundfingerprinting\\perms-new.csv", ",");
Repository repository = new Repository(permutations, databaseService, fingerprintService);
if (scanPath != "") {
if (IOUtils.IsDirectory(scanPath)) {
if (resetdb) {
// For MandelEllis and Scms
mandelEllisScmsDatabase.RemoveTable();
mandelEllisScmsDatabase.AddTable();
// For AudioFingerprinting
databaseService.RemoveFingerprintTable();
databaseService.AddFingerprintTable();
databaseService.RemoveHashBinTable();
databaseService.AddHashBinTable();
databaseService.RemoveTrackTable();
databaseService.AddTrackTable();
}
Console.WriteLine("FindSimilar. Version {0}.", VERSION);
ScanDirectory(scanPath, mandelEllisScmsDatabase, repository, skipDurationAboveSeconds, silent);
} else {
Console.Out.WriteLine("No directory found {0}!", scanPath);
}
}
if (queryPath != "") {
FileInfo fi = new FileInfo(queryPath);
if (fi.Exists) {
FindSimilar(queryPath, mandelEllisScmsDatabase, analysisMethod, numToTake, percentage, distanceType);
} else {
Console.Out.WriteLine("No file found {0}!", queryPath);
}
}
if (queryId != -1) {
FindSimilar(new int[] { queryId }, mandelEllisScmsDatabase, analysisMethod, numToTake, percentage, distanceType);
}
System.Console.ReadLine();
}
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;
}
// TODO: Rememeber to use another stride when querying
public static Dictionary<Track, double> SimilarTracksSoundfingerprinting(FileInfo filePath)
{
DbgTimer t = new DbgTimer();
t.Start ();
FindSimilar.AudioProxies.BassProxy bass = FindSimilar.AudioProxies.BassProxy.Instance;
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;
}
// Name of file being processed
string name = StringUtils.RemoveNonAsciiCharacters(Path.GetFileNameWithoutExtension(filePath.Name));
// 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
// Get database
DatabaseService databaseService = DatabaseService.Instance;
IPermutations permutations = new LocalPermutations("Soundfingerprinting\\perms.csv", ",");
Repository repository = new Repository(permutations, databaseService, fingerprintService);
// work config
WorkUnitParameterObject param = new WorkUnitParameterObject();
param.FingerprintingConfiguration = fingerprintingConfig;
param.PathToAudioFile = filePath.FullName;
param.AudioSamples = audiodata;
param.MillisecondsToProcess = SECONDS_TO_ANALYZE * 1000;
param.StartAtMilliseconds = 0;
Dictionary<Track, double> candidates = repository.FindSimilarFromAudioSamples(25, 4, 2, param);
return candidates;
/*
// Use var keyword to enumerate dictionary
foreach (var pair in candidates)
{
Console.WriteLine("{0} - {1:0.00}",
pair.Key.Title,
pair.Value);
}
*/
Dbg.WriteLine ("Soundfingerprinting - Total Execution Time: {0} ms", t.Stop().TotalMilliseconds);
}
//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 CompareAudioForm()
{
//
// The InitializeComponent() call is required for Windows Forms designer support.
//
InitializeComponent();
//
// TODO: Add constructor code after the InitializeComponent() call.
//
// Instansiate Soundfingerprinting Repository
FingerprintService fingerprintService = Analyzer.GetSoundfingerprintingService();
this.databaseService = DatabaseService.Instance;
IPermutations permutations = new LocalPermutations("Soundfingerprinting\\perms.csv", ",");
//IPermutations permutations = new LocalPermutations("Soundfingerprinting\\perms-new.csv", ",");
IFingerprintingConfiguration fingerprintingConfigCreation = new FullFrequencyFingerprintingConfiguration();
repository = new Repository(permutations, databaseService, fingerprintService);
ImageService imageService = new ImageService(fingerprintService.SpectrumService, fingerprintService.WaveletService);
FileInfo filePathAudio1 = new FileInfo(@"C:\Users\perivar.nerseth\Music\Test Samples Database\VDUB1 Snare 004.wav");
FileInfo filePathAudio2 = new FileInfo(@"C:\Users\perivar.nerseth\Music\Test Samples Search\VDUB1 Snare 004 - Start.wav");
int fingerprintsPerRow = 2;
double[][] logSpectrogram1 = null;
double[][] logSpectrogram2 = null;
List<bool[]> fingerprints1 = null;
List<bool[]> fingerprints2 = null;
WorkUnitParameterObject file1Param = Analyzer.GetWorkUnitParameterObjectFromAudioFile(filePathAudio1);
if (file1Param != null) {
file1Param.FingerprintingConfiguration = fingerprintingConfigCreation;
// Get fingerprints
fingerprints1 = fingerprintService.CreateFingerprintsFromAudioSamples(file1Param.AudioSamples, file1Param, out logSpectrogram1);
pictureBox1.Image = imageService.GetSpectrogramImage(logSpectrogram1, logSpectrogram1.Length, logSpectrogram1[0].Length);
pictureBoxWithInterpolationMode1.Image = imageService.GetImageForFingerprints(fingerprints1, file1Param.FingerprintingConfiguration.FingerprintLength, file1Param.FingerprintingConfiguration.LogBins, fingerprintsPerRow);
}
WorkUnitParameterObject file2Param = Analyzer.GetWorkUnitParameterObjectFromAudioFile(filePathAudio2);
if (file2Param != null) {
file2Param.FingerprintingConfiguration = fingerprintingConfigCreation;
// Get fingerprints
fingerprints2 = fingerprintService.CreateFingerprintsFromAudioSamples(file2Param.AudioSamples, file2Param, out logSpectrogram2);
pictureBox2.Image = imageService.GetSpectrogramImage(logSpectrogram2, logSpectrogram2.Length, logSpectrogram2[0].Length);
pictureBoxWithInterpolationMode2.Image = imageService.GetImageForFingerprints(fingerprints2, file2Param.FingerprintingConfiguration.FingerprintLength, file2Param.FingerprintingConfiguration.LogBins, fingerprintsPerRow);
}
MinHash minHash = repository.MinHash;
// only use the first signatures
bool[] signature1 = fingerprints1[0];
bool[] signature2 = fingerprints2[0];
if (signature1 != null && signature2 != null) {
int hammingDistance = MinHash.CalculateHammingDistance(signature1, signature2);
double jaqSimilarity = MinHash.CalculateJaqSimilarity(signature1, signature2);
lblSimilarity.Text = String.Format("Hamming: {0} JAQ: {1}", hammingDistance, jaqSimilarity);
}
}
/// <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;
}
/// <summary>
/// Query the database for perceptually similar tracks using the sound fingerprinting methods
/// </summary>
/// <param name="filePath">input file</param>
/// <param name="repository">the database (repository)</param>
/// <param name="thresholdTables">Minimum number of hash tables that must be found for one signature to be considered a candidate (0 and 1 = return all candidates, 2+ = return only exact matches)</param>
/// <param name="optimizeSignatureCount">Reduce the number of signatures in order to increase the search performance</param>
/// <param name="doSearchEverything">disregard the local sensitivity hashes and search the whole database</param>
/// <param name="splashScreen">The "please wait" splash screen (or null)</param>
/// <returns>a list of query results objects (e.g. similar tracks)</returns>
public static List<FindSimilar.QueryResult> SimilarTracksSoundfingerprintingList(FileInfo filePath,
Repository repository,
int thresholdTables,
bool optimizeSignatureCount,
bool doSearchEverything,
SplashSceenWaitingForm splashScreen) {
DbgTimer t = new DbgTimer();
t.Start ();
if (splashScreen != null) splashScreen.SetProgress(0, "Reading audio file ...");
// get work config from the audio file
WorkUnitParameterObject param = GetWorkUnitParameterObjectFromAudioFile(filePath);
if (param == null) {
if (splashScreen != null) splashScreen.SetProgress(0, "Failed reading audio file!");
return null;
}
param.FingerprintingConfiguration = fingerprintingConfigQuerying;
if (splashScreen != null) splashScreen.SetProgress(1, "Successfully reading audio file!");
// This is how the threshold tables work:
// For each signature created from a query file we retrieve a number of candidates
// based on how many fingerprints that are associated to the same hash bucket.
// if the number of fingerprints associated to the same hash bucket is relatively high
// the likelyhood for this being an exact match is also very high.
// Therefore a value of 0 or 1 basically means return every track that has an association
// to the same hash bucket, while a number higher than that increases the accuracy for
// only matching identical matches.
// 0 and 1 returns many matches
// 2 returns sometimes only the one we search for (exact match)
List<FindSimilar.QueryResult> similarFiles = repository.FindSimilarFromAudioSamplesList(param.FingerprintingConfiguration.NumberOfHashTables,
param.FingerprintingConfiguration.NumberOfKeys,
thresholdTables,
param,
optimizeSignatureCount,
doSearchEverything,
splashScreen);
Dbg.WriteLine ("SimilarTracksSoundfingerprintingList - Total Execution Time: {0} ms", t.Stop().TotalMilliseconds);
return similarFiles;
}
public FindSimilarClientForm()
{
//
// The InitializeComponent() call is required for Windows Forms designer support.
//
InitializeComponent();
//
// Constructor code after the InitializeComponent() call.
//
this.version.Text = Mirage.Mir.VERSION;
this.DistanceTypeCombo.DataSource = Enum.GetValues(typeof(AudioFeature.DistanceType));
this.ThresholdTablesCombo.DataSource = Enum.GetValues(typeof(ThresholdTables));
// Instansiate SCMS or Mandel Ellis Repository
this.db = new Db();
// Instansiate Soundfingerprinting Repository
FingerprintService fingerprintService = Analyzer.GetSoundfingerprintingService();
this.databaseService = DatabaseService.Instance;
IPermutations permutations = new LocalPermutations("Soundfingerprinting\\perms.csv", ",");
//IPermutations permutations = new LocalPermutations("Soundfingerprinting\\perms-new.csv", ",");
repository = new Repository(permutations, databaseService, fingerprintService);
if (rbScms.Checked) {
IgnoreFileLengthCheckBox.Visible = true;
DistanceTypeCombo.Visible = true;
LessAccurateCheckBox.Visible = false;
ThresholdTablesCombo.Visible = false;
SearchAllFilesCheckbox.Visible = false;
} else {
IgnoreFileLengthCheckBox.Visible = false;
DistanceTypeCombo.Visible = false;
LessAccurateCheckBox.Visible = true;
ThresholdTablesCombo.Visible = true;
SearchAllFilesCheckbox.Visible = true;
}
ReadAllTracks();
}