private void UpdateRow(TrainFilesCarrier.TrainFileRow row)
{
string file = VCDir.Instance.TrainDirAudio + row.Path;
if (File.Exists(file))
{
_yourChanged = true;
_yourPath = file;
_yourWav = new WavFileWrapper(_yourPath);
if (_yourWav.Load())
{
_yourWav.NormalizeWave(1.0);
LogUtil.Info("Load Wave: {0} -- OK\n", _yourPath);
option = VCContext.Instance.MFCCOptions;
if (option.ShiftSampleToZero)
{
LogUtil.Info("Shift Sample To Zero: -- -- OK\n");
_yourWav.ShifToZero();
}
waveViewer.WaveData = _yourWav.FullData;
waveViewer.FitToScreen();
_label = row.Word;
int size = _yourWav.FullData.Count;
_startSelected = (float)row.Start / size;
_endSelected = (float)row.End / size;
waveViewer.LeftSlider = _startSelected;
waveViewer.RightSlider = _endSelected;
}
}
}
public MainTestControlTest()
{
option = VCContext.Instance.MFCCOptions;
//ExtractionWrapper.OptionWrapper.SetLog(option.EnableLog);
InitializeComponent();
selectedChart = showChart.Selected;
}
private void SettingChanged(object obj, SettingEventArgs e)
{
leftChanged = true;
rightChanged = true;
option = e.Option;
//ExtractionWrapper.OptionWrapper.SetLog(option.EnableLog);
}
private void SetValueOnGUI(MfccOptions opt)
{
if (InvokeRequired)
{
Action <MfccOptions> action = new Action <MfccOptions>(SetValueOnGUI);
Invoke(action, new object[] { opt });
return;
}
else
{
//if (options.CepFilter != opt.CepFilter) {
cepfilter_tb.Text = String.Format("{0:0}", opt.CepFilter);
//}
//if (options.NumCeps != opt.NumCeps) {
numceps_tb.Text = String.Format("{0:0}", opt.NumCeps);
//}
//if (options.LowFreq != opt.LowFreq) {
lowfreq_tb.Text = String.Format("{0:0.000}", opt.LowFreq);
//}
//if (options.HighFreq != opt.HighFreq) {
highfreq_tb.Text = String.Format("{0:0.000}", opt.HighFreq);
//}
//if (options.TimeFrame != opt.TimeFrame) {
timeframe_tb.Text = String.Format("{0:0.000}", opt.TimeFrame);
//}
//if (options.TimeShift != opt.TimeShift) {
timeshift_tb.Text = String.Format("{0:0.000}", opt.TimeShift);
//}
}
}
private void SettingChanged(object obj, SettingEventArgs e)
{
_refChanged = true;
_yourChanged = true;
option = e.Option;
ExtractionWrapper.OptionWrapper.SetLog(e.Option.LogLevel);
}
private void SettingChanged(object obj, SettingEventArgs e)
{
refChanged = true;
yourChanged = true;
option = e.Option;
ExtractionWrapper.OptionWrapper.SetLog(option.EnableLog);
}
public MainTestControl()
{
option = new MfccOptions();
ExtractionWrapper.OptionWrapper.SetLog(option.EnableLog);
InitializeComponent();
selectedChart = showChart.Selected;
initListWords();
}
private void Update(MfccOptions opt)
{
this.cepfilter_tb.Text = opt.CepFilter.ToString();
this.highfreq_tb.Text = opt.HighFreq.ToString();
this.numceps_tb.Text = opt.NumCeps.ToString();
this.lowfreq_tb.Text = opt.LowFreq.ToString();
this.timeshift_tb.Text = opt.TimeShift.ToString();
this.timeframe_tb.Text = opt.TimeFrame.ToString();
this.useStandardization_cbx.Checked = opt.UseStandardization;
this.pitchtype_cbx.SelectedIndex = opt.PitchType;
this.yinThreshold_tbx.Text = opt.YinThreshhold.ToString();
this.hightFreq_tbx.Text = opt.PitchHighFreq.ToString();
this.lowFreq_tbx.Text = opt.PitchLowFreq.ToString();
this.timeshift_tbl.Text = opt.PitchTimeShift.ToString();
this.timeframe_tbl.Text = opt.PitchTimeFrame.ToString();
this.median_cb.Checked = opt.UseMedian;
this.median_tbl.Text = opt.MedianWindow.ToString();
this.removeUnpitch_cb.Checked = opt.DropUnPitch;
this.pitch_tbx.Text = opt.PitchThreshold.ToString();
this.energy_txb.Text = opt.EnergyThreshold.ToString();
// Train Tab:
hmmStateNum_tbx.Text = opt.TrainHMMState.ToString();
gmmCompNum_tbx.Text = opt.TrainGMMComponent.ToString();
dataType_cbx.SelectedIndex = (int)opt.TrainCofficientType;
gmmCoVarType_cbx.SelectedIndex = (int)opt.TrainGMMCovVar;
normal_audio_cbx.Checked = opt.NormalizeAudio;
remove_noise_cbx.Checked = opt.RemoveNoiseYourAudio;
shiftToZero_cbx.Checked = opt.ShiftSampleToZero;
// Log
enanblelog_cbx.Checked = opt.EnableLog;
logLevel_cbx.SelectedIndex = 0;
if (opt.LogLevel == (int)LOGLEVEL.STEP)
{
logLevel_cbx.SelectedIndex = 1;
}
else if (opt.LogLevel == (int)LOGLEVEL.INFORMATION)
{
logLevel_cbx.SelectedIndex = 2;
}
else if (opt.LogLevel == (int)LOGLEVEL.DETAIL)
{
logLevel_cbx.SelectedIndex = 3;
}
else if (opt.LogLevel == (int)LOGLEVEL.DATA)
{
logLevel_cbx.SelectedIndex = 4;
}
}
public MainControl()
{
option = VCContext.Instance.MFCCOptions;
//ExtractionWrapper.OptionWrapper.SetLog(option.EnableLog);
recoding = false;
InitializeComponent();
waveViewer.TimeSelectedChanged += SelectedTimeEventHandler;
_waveOut = new NAudio.Wave.WaveOut();
_selectedChart = showChart.Selected;
initSampleRate_cbx();
//initListWords();
FreshListDevices();
}
private VCContext()
{
_watcherListFile = new FileSystemWatcher(VCDir.Instance.XMLDir);
_watcherListFile.NotifyFilter = NotifyFilters.LastWrite;
_watcherListFile.Filter = ".xml";
_watcherListFile.Changed += FileSystemChangeEventHandler;
_watcherListFile.EnableRaisingEvents = true;
option = new MfccOptions();
option.LoadFromXML(VCDir.Instance.SettingFile);
PaserWordTask listWordParser = new PaserWordTask();
if (listWordParser.LoadData(VCDir.Instance.ListWordDir))
{
_listAmTiet = listWordParser.ListAmTiet;
}
}
private void bgw_RunWorkerCompleted(object sender, RunWorkerCompletedEventArgs e)
{
Debug.WriteLine("Complete Procesing: At - {0}", DateTime.Now);
ShowChart();
_refChanged = false;
_yourChanged = false;
process_btn.Enabled = true;
option = VCContext.Instance.MFCCOptions;
if ((bool)e.Result)
{
reg_btn.Enabled = true;
}
if (option.SeparateLog && option.EnableLog && (_refChanged || _yourChanged))
{
ExtractionWrapper.OptionWrapper.SeparateLog();
}
}
public TestControl()
{
locationChart = new Point[8];
locationChart[0] = new Point(Screen.PrimaryScreen.WorkingArea.X + 10, Screen.PrimaryScreen.WorkingArea.Y + 10);
// yourvoice_wave = new WaveViewerForm();
refvoice_wave = new WaveViewerForm();
yourvoice_freq = new MfccChartForm();
refvoice_freq = new MfccChartForm();
locationChart[1] = new Point(locationChart[0].X + yourvoice_freq.Width + 10, locationChart[0].Y);
locationChart[2] = new Point(locationChart[0].X, locationChart[0].Y + yourvoice_freq.Height + 10);
locationChart[3] = new Point(locationChart[0].X + yourvoice_freq.Width + 10, locationChart[0].Y + yourvoice_freq.Height + 10);
yourvoice_mfcc = new MfccChartForm();
refvoice_mfcc = new MfccChartForm();
locationChart[4] = new Point(locationChart[0].X, locationChart[0].Y + 2 * yourvoice_freq.Height + 20);
locationChart[5] = new Point(locationChart[0].X + yourvoice_freq.Width + 10, locationChart[0].Y + 2 * yourvoice_freq.Height + 20);
yourvoice_detal = new MfccChartForm();
refvoice_detal = new MfccChartForm();
locationChart[6] = new Point(locationChart[0].X, locationChart[0].Y + 2 * yourvoice_freq.Height + 20);
locationChart[7] = new Point(locationChart[0].X + yourvoice_freq.Width + 10, locationChart[0].Y + 2 * yourvoice_freq.Height + 20);
yourvoice_double = new MfccChartForm();
refvoice_double = new MfccChartForm();
InitializeComponent();
selectShowChart.ShowChartYourWave += DisplayYourWaveChart;
selectShowChart.ShowChartYourFreq += DisplayYourFreqChart;
selectShowChart.ShowChartYourMfcc += DisplayYourMfccChart;
selectShowChart.ShowChartYourDetal += DisplayYourDetalChart;
selectShowChart.ShowChartYourDouble += DisplayYourDoubleChart;
selectShowChart.ShowChartRefWave += DisplayRefWaveChart;
selectShowChart.ShowChartRefFreq += DisplayRefFreqChart;
selectShowChart.ShowChartRefMfcc += DisplayRefMfccChart;
selectShowChart.ShowChartRefDetal += DisplayRefDetalChart;
selectShowChart.ShowChartRefDouble += DisplayRefDoubleChart;
mfcc_setting = new MfccOptions();
}
public void TestOnlineFeatureExtractor()
{
var mfccOptions = new MfccOptions
{
SamplingRate = 8000,
FeatureCount = 5,
FrameSize = 256,
HopSize = 50,
FilterBankSize = 8
};
var signal = new WhiteNoiseBuilder().OfLength(1000).Build();
var mfccExtractor = new MfccExtractor(mfccOptions);
var mfccVectors = mfccExtractor.ComputeFrom(signal);
var onlineMfccExtractor = new OnlineFeatureExtractor(new MfccExtractor(mfccOptions));
var onlineMfccVectors = new List <float[]>();
var i = 0;
while (i < signal.Length)
{
// emulating online blocks with different sizes:
var size = (i + 1) * 15;
var block = signal.Samples.Skip(i).Take(size).ToArray();
var newVectors = onlineMfccExtractor.ComputeFrom(block);
onlineMfccVectors.AddRange(newVectors);
i += size;
}
var diff = mfccVectors.Zip(onlineMfccVectors, (e, o) => e.Zip(o, (f1, f2) => f1 - f2).Sum());
Assert.That(diff, Is.All.EqualTo(0).Within(1e-7f));
}
private void vadVoice()
{
option = VCContext.Instance.MFCCOptions;
if (_yourWav != null && _yourWav.IsValid)
{
VadWrapper vad = new VadWrapper(_yourWav);
ZeroRateWrapper zrc = new ZeroRateWrapper(_yourWav, 0.02f, 0.01f, true);
vad.UseEnergy(0.015f, 0.01f, true, 3, false);
zrc.Process();
if (vad.Process(option.EnergyThreshold))
{
float begin = 0;
float end = 0;
uint deta = 0;
for (uint i = 0; i < vad.GetSizeOfSegment(); i++)
{
if (vad.GetEndSegment(i) - vad.GetStartSegment(i) > deta)
{
deta = vad.GetEndSegment(i) - vad.GetStartSegment(i);
begin = (float)vad.GetStartSegment(i);
end = (float)vad.GetEndSegment(i);
}
}
if (deta > 0)
{
int size = _yourWav.FullData.Count;
_startSelected = begin / size;
_endSelected = end / size;
waveViewer.LeftSlider = _startSelected;
waveViewer.RightSlider = _endSelected;
waveViewer.ThresholdChart = (float)vad.ThresholdEnergy;
//waveViewer.Chart = vad.SmoothEnergies;
waveViewer.ChartBlue = zrc.ZeroRate;
}
}
}
}
static void Main(string[] args)
{
DiscreteSignal signal;
// load
var mfcc_no = 24;
var samplingRate = 16000;
var mfccOptions = new MfccOptions
{
SamplingRate = samplingRate,
FeatureCount = mfcc_no,
FrameDuration = 0.025 /*sec*/,
HopDuration = 0.010 /*sec*/,
PreEmphasis = 0.97,
Window = WindowTypes.Hamming
};
var opts = new MultiFeatureOptions
{
SamplingRate = samplingRate,
FrameDuration = 0.025,
HopDuration = 0.010
};
var tdExtractor = new TimeDomainFeaturesExtractor(opts);
var mfccExtractor = new MfccExtractor(mfccOptions);
var folders = Directory.GetDirectories(Path.Combine(Environment.CurrentDirectory, "Dataset"));
Console.WriteLine($"Started!");
using (var writer = File.CreateText(Path.Combine(Environment.CurrentDirectory, "Data.csv")))
{
//Write header
var main_header = "genre,";
main_header += String.Join(",", mfccExtractor.FeatureDescriptions);
main_header += ",";
main_header += String.Join(",", tdExtractor.FeatureDescriptions);
main_header += ",centroid,spread,flatness,noiseness,roloff,crest,decrease,spectral_entropy";
writer.WriteLine(main_header);
string feature_string = String.Empty;
foreach (var folder in folders)
{
var f_name = new DirectoryInfo(folder).Name;
var files = Directory.GetFiles(Path.Combine(Environment.CurrentDirectory, "Dataset", folder));
//Write the genre label here
Console.WriteLine($"{f_name}");
foreach (var filename in files)
{
feature_string = String.Empty;
feature_string = $"{f_name},";
//MFCC
var avg_vec_mfcc = new List <float>(mfcc_no + 1);
//TD Features
var avg_vec_td = new List <float>(4);
//Spectral features
var avg_vec_spect = new List <float>(10);
for (var i = 0; i < mfcc_no; i++)
{
avg_vec_mfcc.Add(0f);
}
for (var i = 0; i < 4; i++)
{
avg_vec_td.Add(0f);
}
for (var i = 0; i < 10; i++)
{
avg_vec_spect.Add(0f);
}
string specFeatures = String.Empty;
using (var stream = new FileStream(Path.Combine(Environment.CurrentDirectory, "Dataset", filename), FileMode.Open))
{
var waveFile = new WaveFile(stream);
signal = waveFile[Channels.Average];
//Compute MFCC
tdVectors = tdExtractor.ComputeFrom(signal);
mfccVectors = mfccExtractor.ComputeFrom(signal);
var fftSize = 1024;
var fft = new Fft(fftSize);
var resolution = (float)samplingRate / fftSize;
var frequencies = Enumerable.Range(0, fftSize / 2 + 1)
.Select(f => f * resolution)
.ToArray();
var spectrum = new Fft(fftSize).MagnitudeSpectrum(signal).Samples;
var centroid = Spectral.Centroid(spectrum, frequencies);
var spread = Spectral.Spread(spectrum, frequencies);
var flatness = Spectral.Flatness(spectrum, 0);
var noiseness = Spectral.Noiseness(spectrum, frequencies, 3000);
var rolloff = Spectral.Rolloff(spectrum, frequencies, 0.85f);
var crest = Spectral.Crest(spectrum);
var decrease = Spectral.Decrease(spectrum);
var entropy = Spectral.Entropy(spectrum);
specFeatures = $"{centroid},{spread},{flatness},{noiseness},{rolloff},{crest},{decrease},{entropy}";
}
//Write label here TODO
foreach (var inst in mfccVectors)
{
for (var i = 0; i < mfcc_no; i++)
{
avg_vec_mfcc[i] += inst[i];
}
}
foreach (var inst in tdVectors)
{
for (var i = 0; i < 4; i++)
{
avg_vec_td[i] += inst[i];
}
}
for (var i = 0; i < mfcc_no; i++)
{
avg_vec_mfcc[i] /= mfccVectors.Count;
}
for (var i = 0; i < 4; i++)
{
avg_vec_td[i] /= tdVectors.Count;
}
// Write MFCCs
feature_string += String.Join(",", avg_vec_mfcc);
feature_string += ",";
feature_string += String.Join(",", avg_vec_td);
//Write Spectral features as well
feature_string += ",";
feature_string += specFeatures;
writer.WriteLine(feature_string);
var file_name = new DirectoryInfo(filename).Name;
Console.WriteLine($"{file_name}");
}
}
}
Console.WriteLine($"DONE");
Console.ReadLine();
}
public MfccSetting()
{
options = new MfccOptions();
options.Reset();
InitializeComponent();
}
public MfccExtractorTestHtk(MfccOptions options) : base(options)
{
_hammingWin = Window.OfType(WindowTypes.Hamming, FrameSize);
}
/// <summary>
/// Constructs extractor from configuration <paramref name="options"/>.
/// </summary>
public MfccExtractor(MfccOptions options) : base(options)
{
FeatureCount = options.FeatureCount;
var filterbankSize = options.FilterBankSize;
if (options.FilterBank is null)
{
_blockSize = options.FftSize > FrameSize ? options.FftSize : MathUtils.NextPowerOfTwo(FrameSize);
var melBands = FilterBanks.MelBands(filterbankSize, SamplingRate, options.LowFrequency, options.HighFrequency);
FilterBank = FilterBanks.Triangular(_blockSize, SamplingRate, melBands, mapper: Scale.HerzToMel); // HTK/Kaldi-style
}
else
{
FilterBank = options.FilterBank;
filterbankSize = FilterBank.Length;
_blockSize = 2 * (FilterBank[0].Length - 1);
Guard.AgainstExceedance(FrameSize, _blockSize, "frame size", "FFT size");
}
_fft = new RealFft(_blockSize);
_lifterSize = options.LifterSize;
_lifterCoeffs = _lifterSize > 0 ? Window.Liftering(FeatureCount, _lifterSize) : null;
_includeEnergy = options.IncludeEnergy;
_logEnergyFloor = options.LogEnergyFloor;
// setup DCT: ============================================================================
_dctType = options.DctType;
switch (_dctType[0])
{
case '1': _dct = new Dct1(filterbankSize); break;
case '3': _dct = new Dct3(filterbankSize); break;
case '4': _dct = new Dct4(filterbankSize); break;
default: _dct = new Dct2(filterbankSize); break;
}
if (_dctType.EndsWith("N", StringComparison.OrdinalIgnoreCase))
{
_applyDct = mfccs => _dct.DirectNorm(_melSpectrum, mfccs);
}
else
{
_applyDct = mfccs => _dct.Direct(_melSpectrum, mfccs);
}
// setup spectrum post-processing: =======================================================
_logFloor = options.LogFloor;
_nonLinearityType = options.NonLinearity;
switch (_nonLinearityType)
{
case NonLinearityType.Log10:
_postProcessSpectrum = () => FilterBanks.ApplyAndLog10(FilterBank, _spectrum, _melSpectrum, _logFloor); break;
case NonLinearityType.LogE:
_postProcessSpectrum = () => FilterBanks.ApplyAndLog(FilterBank, _spectrum, _melSpectrum, _logFloor); break;
case NonLinearityType.ToDecibel:
_postProcessSpectrum = () => FilterBanks.ApplyAndToDecibel(FilterBank, _spectrum, _melSpectrum, _logFloor); break;
case NonLinearityType.CubicRoot:
_postProcessSpectrum = () => FilterBanks.ApplyAndPow(FilterBank, _spectrum, _melSpectrum, 0.33); break;
default:
_postProcessSpectrum = () => FilterBanks.Apply(FilterBank, _spectrum, _melSpectrum); break;
}
_spectrumType = options.SpectrumType;
switch (_spectrumType)
{
case SpectrumType.Magnitude:
_getSpectrum = block => _fft.MagnitudeSpectrum(block, _spectrum, false); break;
case SpectrumType.MagnitudeNormalized:
_getSpectrum = block => _fft.MagnitudeSpectrum(block, _spectrum, true); break;
case SpectrumType.PowerNormalized:
_getSpectrum = block => _fft.PowerSpectrum(block, _spectrum, true); break;
default:
_getSpectrum = block => _fft.PowerSpectrum(block, _spectrum, false); break;
}
// reserve memory for reusable blocks
_spectrum = new float[_blockSize / 2 + 1];
_melSpectrum = new float[filterbankSize];
}
private MfccOptions Get()
{
MfccOptions options = new MfccOptions();
try
{
//MFCC
if (cepfilter_tb.Text.Length > 0)
{
options.CepFilter = Convert.ToUInt32(cepfilter_tb.Text);
}
if (numceps_tb.Text.Length > 0)
{
options.NumCeps = Convert.ToUInt32(numceps_tb.Text);
}
if (lowfreq_tb.Text.Length > 0)
{
options.LowFreq = Convert.ToSingle(lowfreq_tb.Text);
}
if (highfreq_tb.Text.Length > 0)
{
options.HighFreq = Convert.ToSingle(highfreq_tb.Text);
}
if (timeframe_tb.Text.Length > 0)
{
options.TimeFrame = Convert.ToSingle(timeframe_tb.Text);
}
if (timeshift_tb.Text.Length > 0)
{
options.TimeShift = Convert.ToSingle(timeshift_tb.Text);
}
options.UseStandardization = useStandardization_cbx.Checked;
//Pitch
options.PitchType = pitchtype_cbx.SelectedIndex;
if (yinThreshold_tbx.Text.Length > 0)
{
options.YinThreshhold = Convert.ToSingle(yinThreshold_tbx.Text);
}
if (hightFreq_tbx.Text.Length > 0)
{
options.PitchHighFreq = Convert.ToSingle(hightFreq_tbx.Text);
}
if (timeshift_tbl.Text.Length > 0)
{
options.PitchTimeShift = Convert.ToSingle(timeshift_tbl.Text);
}
if (timeframe_tbl.Text.Length > 0)
{
options.PitchTimeFrame = Convert.ToSingle(timeframe_tbl.Text);
}
if (lowFreq_tbx.Text.Length > 0)
{
options.PitchLowFreq = Convert.ToSingle(lowFreq_tbx.Text);
}
options.UseMedian = median_cb.Checked;
if (median_tbl.Text.Length > 0)
{
options.MedianWindow = Convert.ToInt32(median_tbl.Text);
}
options.DropUnPitch = removeUnpitch_cb.Checked;
// VAD
if (energy_txb.Text.Length > 0)
{
options.EnergyThreshold = Convert.ToSingle(energy_txb.Text);
}
if (pitch_tbx.Text.Length > 0)
{
options.PitchThreshold = Convert.ToSingle(pitch_tbx.Text);
}
//Noise and Normalize
options.NormalizeAudio = normal_audio_cbx.Checked;
options.RemoveNoiseYourAudio = remove_noise_cbx.Checked;
options.ShiftSampleToZero = shiftToZero_cbx.Checked;
// Log
options.EnableLog = enanblelog_cbx.Checked;
int selectedText = logLevel_cbx.SelectedIndex;
options.LogLevel = (int)LOGLEVEL.NONE;
if (selectedText == 1)
{
options.LogLevel = (int)LOGLEVEL.STEP;
}
else if (selectedText == 2)
{
options.LogLevel = (int)LOGLEVEL.INFORMATION;
}
else if (selectedText == 3)
{
options.LogLevel = (int)LOGLEVEL.DETAIL;
}
else if (selectedText == 4)
{
options.LogLevel = (int)LOGLEVEL.DATA;
}
//Train
options.TrainHMMState = (uint)Convert.ToInt32(hmmStateNum_tbx.Text);
options.TrainGMMComponent = (uint)Convert.ToInt32(gmmCompNum_tbx.Text);
options.TrainCofficientType = (uint)dataType_cbx.SelectedIndex;
options.TrainGMMCovVar = (uint)gmmCoVarType_cbx.SelectedIndex;
}
catch (Exception)
{
}
return(options);
}
public SettingEventArgs(MfccOptions opt)
{
Option = opt;
}
public SettingChangedEventArgs(MfccOptions value)
{
Value = value;
}
private void showWaveSound()
{
_yourWav = new WavFileWrapper(_yourPath);
if (_yourWav.Load())
{
waveViewer.WaveData = _yourWav.FullData;
waveViewer.FitToScreen();
vadVoice();
if (_regMode)
{
uint size = (uint)_yourWav.FullData.Count;
uint begin = (uint)(_startSelected * size);
uint end = (uint)(_endSelected * size);
_yourWav.NormalizeWave(1.0f);
option = VCContext.Instance.MFCCOptions;
LogUtil.Info("Load Wave: {0} -- OK\n", _yourPath);
if (option.ShiftSampleToZero)
{
LogUtil.Info("Shift Sample To Zero: -- -- OK\n");
_yourWav.ShifToZero();
}
Debug.WriteLine("Select Data voice: Start {0} End {1}", begin, end);
_yourWav.SelectedWave(begin, end);
_yourMfcc = new MFCCWrapper(_yourWav, option.TimeFrame, option.TimeShift, option.CepFilter, option.LowFreq, option.HighFreq, option.NumCeps, 4);
_yourMfcc.UserStandardization = option.UseStandardization;
if (_yourMfcc != null && _yourMfcc.Process())
{
List <List <double> > data = null;
switch (VCContext.Instance.MFCCOptions.TrainCofficientType)
{
case 0:
data = _yourMfcc.Mfcc;
break;
case 1:
data = _yourMfcc.DetalMfcc;
break;
case 2:
data = _yourMfcc.DoubleDetalMfcc;
break;
}
Action act = new Action(() =>
{
reg_lb.Text = _trainTask.Reg(data);
});
Invoke(act);
}
else
{
MessageBox.Show(" Cant Extraction file {0}\n", _yourPath);
}
}
}
}
public void extractFeatures()
{
//NWaves
//Initial setup
if (_filePath != null)
{
DiscreteSignal signal;
// load
var mfcc_no = 24;
var samplingRate = 44100;
var mfccOptions = new MfccOptions
{
SamplingRate = samplingRate,
FeatureCount = mfcc_no,
FrameDuration = 0.025 /*sec*/,
HopDuration = 0.010 /*sec*/,
PreEmphasis = 0.97,
Window = WindowTypes.Hamming
};
var opts = new MultiFeatureOptions
{
SamplingRate = samplingRate,
FrameDuration = 0.025,
HopDuration = 0.010
};
var tdExtractor = new TimeDomainFeaturesExtractor(opts);
var mfccExtractor = new MfccExtractor(mfccOptions);
// Read from file.
featureString = String.Empty;
featureString = $"green,";
//MFCC
var avg_vec_mfcc = new List <float>(mfcc_no + 1);
//TD Features
var avg_vec_td = new List <float>(4);
//Spectral features
var avg_vec_spect = new List <float>(10);
for (var i = 0; i < mfcc_no; i++)
{
avg_vec_mfcc.Add(0f);
}
for (var i = 0; i < 4; i++)
{
avg_vec_td.Add(0f);
}
for (var i = 0; i < 10; i++)
{
avg_vec_spect.Add(0f);
}
string specFeatures = String.Empty;
Console.WriteLine($"{tag} Reading from file");
using (var stream = new FileStream(_filePath, FileMode.Open))
{
var waveFile = new WaveFile(stream);
signal = waveFile[channel : Channels.Left];
////Compute MFCC
float[] mfvfuck = new float[25];
var sig_sam = signal.Samples;
mfccVectors = mfccExtractor.ComputeFrom(sig_sam);
var fftSize = 1024;
tdVectors = tdExtractor.ComputeFrom(signal.Samples);
var fft = new Fft(fftSize);
var resolution = (float)samplingRate / fftSize;
var frequencies = Enumerable.Range(0, fftSize / 2 + 1)
.Select(f => f * resolution)
.ToArray();
var spectrum = new Fft(fftSize).MagnitudeSpectrum(signal).Samples;
var centroid = Spectral.Centroid(spectrum, frequencies);
var spread = Spectral.Spread(spectrum, frequencies);
var flatness = Spectral.Flatness(spectrum, 0);
var noiseness = Spectral.Noiseness(spectrum, frequencies, 3000);
var rolloff = Spectral.Rolloff(spectrum, frequencies, 0.85f);
var crest = Spectral.Crest(spectrum);
var decrease = Spectral.Decrease(spectrum);
var entropy = Spectral.Entropy(spectrum);
specFeatures = $"{centroid},{spread},{flatness},{noiseness},{rolloff},{crest},{decrease},{entropy}";
//}
Console.WriteLine($"{tag} All features ready");
for (int calibC = 0; calibC < mfccVectors.Count; calibC += (mfccVectors.Count / duration) - 1)
{
featureString = String.Empty;
var tmp = new ModelInput();
for (var i = 0; i < mfcc_no; i++)
{
avg_vec_mfcc[i] = mfccVectors[calibC][i];
}
for (var i = 0; i < 4; i++)
{
avg_vec_td[i] = tdVectors[calibC][i];
}
featureString += String.Join(",", avg_vec_mfcc);
featureString += ",";
featureString += String.Join(",", avg_vec_td);
featureString += ",";
featureString += specFeatures;
Console.WriteLine($"{tag} Feature String ready {featureString}");
if (File.Exists(Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.LocalApplicationData), "temp")))
{
File.Delete(Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.LocalApplicationData), "temp"));
File.WriteAllText(Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.LocalApplicationData), "temp"), featureString);
}
else
{
File.WriteAllText(Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.LocalApplicationData), "temp"), featureString);
}
MLContext mLContext = new MLContext();
string fileName = Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.LocalApplicationData), "temp");
IDataView dataView = mLContext.Data.LoadFromTextFile <ModelInput>(
path: fileName,
hasHeader: false,
separatorChar: ',',
allowQuoting: true,
allowSparse: false);
// Use first line of dataset as model input
// You can replace this with new test data (hardcoded or from end-user application)
ModelInput sampleForPrediction = mLContext.Data.CreateEnumerable <ModelInput>(dataView, false)
.First();
ModelOutput opm = ConsumeModel.Predict(sampleForPrediction);
featureTimeList.Add(opm.Score);
Console.WriteLine($"{tag} Feature vs time list ready");
}
//Console.WriteLine($"{tag} MFCC: {mfccVectors.Count}");
//Console.WriteLine($"{tag} TD: {tdVectors.Count}");
//Console.WriteLine($"{tag} featureTimeArray: {featureTimeList.Count} {featureString}");
}
}
}
async public void extractFeatures(string _filepath, StorageFile sf)
{
op = new float[10];
tdVectors = new List <float[]>();
mfccVectors = new List <float[]>();
featureTimeList = new List <float[]>();
//NWaves
FilePath = _filepath;
PredictedLabel = "Ready!.";
//player.Load(GetStreamFromFile(FilePath));
//player.Play();
mMedia.Source = MediaSource.CreateFromStorageFile(sf);
bool test = player.IsPlaying;
mMedia.AutoPlay = true;
MusicProperties properties = await sf.Properties.GetMusicPropertiesAsync();
TimeSpan myTrackDuration = properties.Duration;
duration = Convert.ToInt32(myTrackDuration.TotalSeconds);
if (FilePath != null)
{
DiscreteSignal signal;
// load
var mfcc_no = 24;
var samplingRate = 44100;
var mfccOptions = new MfccOptions
{
SamplingRate = samplingRate,
FeatureCount = mfcc_no,
FrameDuration = 0.025 /*sec*/,
HopDuration = 0.010 /*sec*/,
PreEmphasis = 0.97,
Window = WindowTypes.Hamming
};
var opts = new MultiFeatureOptions
{
SamplingRate = samplingRate,
FrameDuration = 0.025,
HopDuration = 0.010
};
var tdExtractor = new TimeDomainFeaturesExtractor(opts);
var mfccExtractor = new MfccExtractor(mfccOptions);
// Read from file.
featureString = String.Empty;
featureString = $"green,";
//MFCC
var mfccList = new List <List <double> >();
var tdList = new List <List <double> >();
//MFCC
//TD Features
//Spectral features
for (var i = 0; i < mfcc_no; i++)
{
mfccList.Add(new List <double>());
}
for (var i = 0; i < 4; i++)
{
tdList.Add(new List <double>());
}
string specFeatures = String.Empty;
Console.WriteLine($"{tag} Reading from file");
using (var stream = new FileStream(FilePath, FileMode.Open))
{
var waveFile = new WaveFile(stream);
signal = waveFile[channel : Channels.Left];
////Compute MFCC
float[] mfvfuck = new float[25];
var sig_sam = signal.Samples;
mfccVectors = mfccExtractor.ComputeFrom(sig_sam);
var fftSize = 1024;
tdVectors = tdExtractor.ComputeFrom(signal.Samples);
var fft = new Fft(fftSize);
var resolution = (float)samplingRate / fftSize;
var frequencies = Enumerable.Range(0, fftSize / 2 + 1)
.Select(f => f * resolution)
.ToArray();
var spectrum = new Fft(fftSize).MagnitudeSpectrum(signal).Samples;
var centroid = Spectral.Centroid(spectrum, frequencies);
var spread = Spectral.Spread(spectrum, frequencies);
var flatness = Spectral.Flatness(spectrum, 0);
var noiseness = Spectral.Noiseness(spectrum, frequencies, 3000);
var rolloff = Spectral.Rolloff(spectrum, frequencies, 0.85f);
var crest = Spectral.Crest(spectrum);
var decrease = Spectral.Decrease(spectrum);
var entropy = Spectral.Entropy(spectrum);
specFeatures = $"{centroid},{spread},{flatness},{noiseness},{rolloff},{crest},{decrease},{entropy}";
//}
Console.WriteLine($"{tag} All features ready");
for (int calibC = 0; calibC < mfccVectors.Count;)
{
featureString = String.Empty;
var tmp = new ModelInput();
for (var j = 0; j < (mfccVectors.Count / duration) - 1 && calibC < mfccVectors.Count; j++)
{
for (var i = 0; i < mfcc_no; i++)
{
mfccList[i].Add(mfccVectors[calibC][i]);
}
for (var i = 0; i < 4; i++)
{
tdList[i].Add(tdVectors[calibC][i]);
}
calibC += 1;
}
var mfcc_statistics = new List <double>();
for (var i = 0; i < mfcc_no; i++)
{
//preheader += m + "_mean";
//preheader += m + "_min";
//preheader += m + "_var";
//preheader += m + "_sd";
//preheader += m + "_med";
//preheader += m + "_lq";
//preheader += m + "_uq";
//preheader += m + "_skew";
//preheader += m + "_kurt";
mfcc_statistics.Add(Statistics.Mean(mfccList[i]));
mfcc_statistics.Add(Statistics.Minimum(mfccList[i]));
mfcc_statistics.Add(Statistics.Variance(mfccList[i]));
mfcc_statistics.Add(Statistics.StandardDeviation(mfccList[i]));
mfcc_statistics.Add(Statistics.Median(mfccList[i]));
mfcc_statistics.Add(Statistics.LowerQuartile(mfccList[i]));
mfcc_statistics.Add(Statistics.UpperQuartile(mfccList[i]));
mfcc_statistics.Add(Statistics.Skewness(mfccList[i]));
mfcc_statistics.Add(Statistics.Kurtosis(mfccList[i]));
}
var td_statistics = new List <double>();
for (var i = 0; i < 4; i++)
{
td_statistics.Add(Statistics.Mean(tdList[i]));
td_statistics.Add(Statistics.Minimum(tdList[i]));
td_statistics.Add(Statistics.Variance(tdList[i]));
td_statistics.Add(Statistics.StandardDeviation(tdList[i]));
td_statistics.Add(Statistics.Median(tdList[i]));
td_statistics.Add(Statistics.LowerQuartile(tdList[i]));
td_statistics.Add(Statistics.UpperQuartile(tdList[i]));
td_statistics.Add(Statistics.Skewness(tdList[i]));
td_statistics.Add(Statistics.Kurtosis(tdList[i]));
}
// Write MFCCs
featureString += String.Join(",", mfcc_statistics);
featureString += ",";
featureString += String.Join(",", td_statistics);
//Write Spectral features as well
featureString += ",";
featureString += specFeatures;
Console.WriteLine($"{tag} Feature String ready {featureString}");
if (File.Exists(Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.LocalApplicationData), "temp")))
{
File.Delete(Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.LocalApplicationData), "temp"));
File.WriteAllText(Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.LocalApplicationData), "temp"), featureString);
}
else
{
File.WriteAllText(Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.LocalApplicationData), "temp"), featureString);
}
MLContext mLContext = new MLContext();
string fileName = Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.LocalApplicationData), "temp");
IDataView dataView = mLContext.Data.LoadFromTextFile <ModelInput>(
path: fileName,
hasHeader: false,
separatorChar: ',',
allowQuoting: true,
allowSparse: false);
// Use first line of dataset as model input
// You can replace this with new test data (hardcoded or from end-user application)
ModelInput sampleForPrediction = mLContext.Data.CreateEnumerable <ModelInput>(dataView, false)
.First();
ModelOutput opm = ConsumeModel.Predict(sampleForPrediction);
featureTimeList.Add(opm.Score);
Console.WriteLine($"{tag} Feature vs time list ready");
}
//Console.WriteLine($"{tag} MFCC: {mfccVectors.Count}");
//Console.WriteLine($"{tag} TD: {tdVectors.Count}");
//Console.WriteLine($"{tag} featureTimeArray: {featureTimeList.Count} {featureString}");
}
}
playAudio();
}
private void bgw_DoWork(object sender, DoWorkEventArgs e)
{
bool reg = (bool)e.Argument;
if (_yourChanged)
{
_yourWav = new WavFileWrapper(_yourPath);
if (_yourWav.Load())
{
_yourWav.NormalizeWave(1.0);
option = VCContext.Instance.MFCCOptions;
LogUtil.Info("Load Wave: {0} -- OK\n", _yourPath);
if (option.ShiftSampleToZero)
{
LogUtil.Info("Shift Sample To Zero: -- -- OK\n");
_yourWav.ShifToZero();
}
int size = _yourWav.FullData.Count;
uint startPnt = (uint)(_startSelected * size);
uint endPnt = (uint)(_endSelected * size);
Debug.WriteLine("Select Data voice: Start {0} End {1}", startPnt, endPnt);
_yourWav.SelectedWave(startPnt, endPnt);
_yourMfcc = new MFCCWrapper(_yourWav, option.TimeFrame, option.TimeShift, option.CepFilter, option.LowFreq, option.HighFreq, option.NumCeps, 4);
_yourMfcc.UserStandardization = option.UseStandardization;
_yourMfcc.Process();
_yourPitch = new PitchWrapper(_yourWav, option.PitchTimeFrame, option.PitchTimeShift, option.PitchLowFreq, option.PitchHighFreq, option.PitchType, option.DropUnPitch);
if (option.UseMedian)
{
_yourPitch.SetMedianWindowSize(option.MedianWindow);
}
_yourPitch.Process();
}
}
if (_yourMfcc != null && _yourMfcc.ProcessDone)
{
// TO DO: Process Bar
SetDataChart(FormTag.YOUR_WAVE);
SetDataChart(FormTag.YOUR_MFCC);
SetDataChart(FormTag.YOUR_FREQ);
SetDataChart(FormTag.YOUR_DOUBLE);
SetDataChart(FormTag.YOUR_DETAL);
SetDataChart(FormTag.YOUR_PITCH);
}
e.Result = reg;
if (reg)
{
if (_yourMfcc != null)
{
List <List <double> > data = null;
switch (VCContext.Instance.MFCCOptions.TrainCofficientType)
{
case 0:
data = _yourMfcc.Mfcc;
break;
case 1:
data = _yourMfcc.DetalMfcc;
break;
case 2:
data = _yourMfcc.DoubleDetalMfcc;
break;
}
Action act = new Action(() =>
{
reg_lb.Text = _trainTask.Reg(data);
});
Invoke(act);
}
}
}