FormatClassifier classifies the sample format and endianness of raw audio files. More...

#include <FormatClassifier.h>

Collaboration diagram for FormatClassifier:

Classes
struct	FormatClassT

Public Types
typedef std::vector< FormatClassT >	FormatVectorT

typedef std::vector< FormatClassT >::iterator	FormatVectorIt

Public Member Functions
	FormatClassifier (const char *filename)

	~FormatClassifier ()

FormatClassT	GetResultFormat ()

int	GetResultFormatLibSndfile ()

unsigned	GetResultChannels ()

Private Member Functions
void	Run ()

void	ReadSignal (FormatClassT format, size_t stride)

void	ConvertSamples (void in, float out, FormatClassT format)

void	Add (float in1, float in2, size_t len)

void	Sub (float *in, float subt, size_t len)

void	Div (float *in, float div, size_t len)

void	Abs (float in, float out, size_t len)

float	Mean (float *in, size_t len)

float	Max (float *in, size_t len)

float	Max (float in, size_t len, size_t maxidx)

template<class T >
void	ToFloat (T in, float out, size_t len)

Private Attributes
FormatVectorT	mClasses

MultiFormatReader	mReader

SpecPowerCalculation	mMeter

Floats	mSigBuffer { cSiglen }

Floats	mAuxBuffer { cSiglen }

ArrayOf< uint8_t >	mRawBuffer { cSiglen * 8 }

Floats	mMonoFeat

Floats	mStereoFeat

FormatClassT	mResultFormat

unsigned	mResultChannels { 0 }

Static Private Attributes
static const size_t	cSiglen = 512

static const size_t	cNumInts = 32

Detailed Description

FormatClassifier classifies the sample format and endianness of raw audio files.

The classifier operates in the frequency domain and exploits the low-pass-like spectral behaviour of natural audio signals for classification of the sample format and the used endianness.

Definition at line 56 of file FormatClassifier.h.

Member Typedef Documentation

◆ FormatVectorIt

typedef std::vector<FormatClassT>::iterator FormatClassifier::FormatVectorIt

Definition at line 67 of file FormatClassifier.h.

◆ FormatVectorT

typedef std::vector<FormatClassT> FormatClassifier::FormatVectorT

Definition at line 66 of file FormatClassifier.h.

Constructor & Destructor Documentation

◆ FormatClassifier()

FormatClassifier::FormatClassifier ( const char * filename )

Definition at line 31 of file FormatClassifier.cpp.

                                                       :
   mReader(filename),
   mMeter(cSiglen)
{
   // Define the classification classes
   for ( auto endianness : {
      MachineEndianness::Little,
      MachineEndianness::Big,
   } )
      for ( auto format : {
         MultiFormatReader::Int8,
         MultiFormatReader::Int16,
         MultiFormatReader::Int32,
         MultiFormatReader::Uint8,
         MultiFormatReader::Float,
         MultiFormatReader::Double,
      } )
         mClasses.push_back( { format, endianness } );
 
   // Build feature vectors
   mMonoFeat = Floats{ mClasses.size() };
   mStereoFeat = Floats{ mClasses.size() };
   
#ifdef FORMATCLASSIFIER_SIGNAL_DEBUG
   // Build a debug writer
   char dfile [1024];
   sprintf(dfile, "%s.sig", filename);
   mpWriter = std::make_unique<DebugWriter>(dfile);
#endif
 
   // Run it
   Run();
   
#ifdef FORMATCLASSIFIER_SIGNAL_DEBUG
   for (unsigned int n = 0; n < mClasses.size(); n++)
   {
      wxPrintf("Class [%i] Machine [%i]: Mono: %3.7f Stereo: %3.7f\n", mClasses[n].format, mClasses[n].endian, mMonoFeat[n], mStereoFeat[n]);
   }
#endif
 
}

References MachineEndianness::Big, MultiFormatReader::Double, MultiFormatReader::Float, anonymous_namespace{ExportPCM.cpp}::format, MultiFormatReader::Int16, MultiFormatReader::Int32, MultiFormatReader::Int8, MachineEndianness::Little, mClasses, mMonoFeat, mStereoFeat, Run(), and MultiFormatReader::Uint8.

Here is the call graph for this function:

◆ ~FormatClassifier()

FormatClassifier::~FormatClassifier ( )

Definition at line 73 of file FormatClassifier.cpp.

74{

75}

Member Function Documentation

◆ Abs()

void FormatClassifier::Abs	(	float *	in,
		float *	out,
		size_t	len
	)

private

Definition at line 293 of file FormatClassifier.cpp.

{
   for (unsigned int n = 0; n < len; n++)
   {
      if (in[n] < 0.0f)
      {
         out[n] = -in[n];
      }
      else
      {
         out[n] = in[n];
      }
   }
}

Referenced by Run().

Here is the caller graph for this function:

◆ Add()

void FormatClassifier::Add	(	float *	in1,
		float *	in2,
		size_t	len
	)

private

Definition at line 268 of file FormatClassifier.cpp.

{
   for (unsigned int n = 0; n < len; n++)
   {
      in1[n] += in2[n];
   }
}

Referenced by ReadSignal().

Here is the caller graph for this function:

◆ ConvertSamples()

void FormatClassifier::ConvertSamples	(	void *	in,
		float *	out,
		FormatClassT	format
	)

private

Definition at line 237 of file FormatClassifier.cpp.

{
   switch(format.format)
   {
      case MultiFormatReader::Int8:
         ToFloat((int8_t*) in, out, cSiglen);
         break;
      case MultiFormatReader::Int16:
         ToFloat((int16_t*) in, out, cSiglen);
         break;
      case MultiFormatReader::Int32:
         ToFloat((int32_t*) in, out, cSiglen);
         break;
      case MultiFormatReader::Uint8:
         ToFloat((uint8_t*) in, out, cSiglen);
         break;
      case MultiFormatReader::Uint16:
         ToFloat((uint16_t*) in, out, cSiglen);
         break;
      case MultiFormatReader::Uint32:
         ToFloat((uint32_t*) in, out, cSiglen);
        break;
      case MultiFormatReader::Float:
         ToFloat((float*) in, out, cSiglen);
         break;
      case MultiFormatReader::Double:
         ToFloat((double*) in, out, cSiglen);
         break;
   }
}

References cSiglen, MultiFormatReader::Double, MultiFormatReader::Float, anonymous_namespace{ExportPCM.cpp}::format, MultiFormatReader::Int16, MultiFormatReader::Int32, MultiFormatReader::Int8, ToFloat(), MultiFormatReader::Uint16, MultiFormatReader::Uint32, and MultiFormatReader::Uint8.

Referenced by ReadSignal().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ Div()

void FormatClassifier::Div	(	float *	in,
		float	div,
		size_t	len
	)

private

Definition at line 284 of file FormatClassifier.cpp.

{
   for (unsigned int n = 0; n < len; n++)
   {
      in[n] /= div;
   }
}

Referenced by Run().

Here is the caller graph for this function:

◆ GetResultChannels()

unsigned FormatClassifier::GetResultChannels ( )

Definition at line 124 of file FormatClassifier.cpp.

{
   return mResultChannels;
}

References mResultChannels.

Referenced by ImportRawDialog::OnDetect().

Here is the caller graph for this function:

◆ GetResultFormat()

FormatClassifier::FormatClassT FormatClassifier::GetResultFormat ( )

Definition at line 77 of file FormatClassifier.cpp.

{
   return mResultFormat;
}

References mResultFormat.

◆ GetResultFormatLibSndfile()

int FormatClassifier::GetResultFormatLibSndfile ( )

Definition at line 82 of file FormatClassifier.cpp.

{
   int format = SF_FORMAT_RAW;
   
   switch(mResultFormat.format)
   {
      case MultiFormatReader::Int8:
         format |= SF_FORMAT_PCM_S8;
         break;
      case MultiFormatReader::Int16:
         format |= SF_FORMAT_PCM_16;
         break;
      case MultiFormatReader::Int32:
         format |= SF_FORMAT_PCM_32;
         break;
      case MultiFormatReader::Uint8:
         format |= SF_FORMAT_PCM_U8;
         break;
      case MultiFormatReader::Float:
         format |= SF_FORMAT_FLOAT;
         break;
      case MultiFormatReader::Double:
         format |= SF_FORMAT_DOUBLE;
         break;
      default:
         format |= SF_FORMAT_PCM_16;
         break;
   }
   
   switch(mResultFormat.endian)
   {
      case MachineEndianness::Little:
         format |= SF_ENDIAN_LITTLE;
         break;
      case MachineEndianness::Big:
         format |= SF_ENDIAN_BIG;
         break;
   }
   
   return format;
}

References MachineEndianness::Big, MultiFormatReader::Double, FormatClassifier::FormatClassT::endian, MultiFormatReader::Float, anonymous_namespace{ExportPCM.cpp}::format, FormatClassifier::FormatClassT::format, MultiFormatReader::Int16, MultiFormatReader::Int32, MultiFormatReader::Int8, MachineEndianness::Little, mResultFormat, and MultiFormatReader::Uint8.

Referenced by ImportRawDialog::OnDetect().

Here is the caller graph for this function:

◆ Max() [1/2]

float FormatClassifier::Max	(	float *	in,
		size_t	len
	)

private

Definition at line 322 of file FormatClassifier.cpp.

{
   size_t dummyidx;
   return Max(in, len, &dummyidx);
}

References Max().

Referenced by Max(), and Run().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ Max() [2/2]

float FormatClassifier::Max	(	float *	in,
		size_t	len,
		size_t *	maxidx
	)

private

Definition at line 328 of file FormatClassifier.cpp.

{
   float max = -FLT_MAX;
   *maxidx = 0;
   
   for (unsigned int n = 0; n < len; n++)
   {
      if (in[n] > max)
      {
         max = in[n];
         *maxidx = n;
      }
   }
 
   return max;
}

◆ Mean()

float FormatClassifier::Mean	(	float *	in,
		size_t	len
	)

private

Definition at line 308 of file FormatClassifier.cpp.

{
   float mean = 0.0f;
 
   for (unsigned int n = 0; n < len; n++)
   {
      mean += in[n];
   }
 
   mean /= len;
   
   return mean;
}

Referenced by Run().

Here is the caller graph for this function:

◆ ReadSignal()

void FormatClassifier::ReadSignal	(	FormatClassT	format,
		size_t	stride
	)

private

Definition at line 199 of file FormatClassifier.cpp.

{
   size_t actRead = 0;
   unsigned int n = 0;
 
   mReader.Reset();
 
   // Do a dummy read of 1024 bytes to skip potential header information
   mReader.ReadSamples(mRawBuffer.get(), 1024, MultiFormatReader::Uint8, MachineEndianness::Little);
 
   do
   {
      actRead = mReader.ReadSamples(mRawBuffer.get(), cSiglen, stride, format.format, format.endian);
 
      if (n == 0)
      {
         ConvertSamples(mRawBuffer.get(), mSigBuffer.get(), format);
      }
      else
      {
         if (actRead == cSiglen)
         {
            ConvertSamples(mRawBuffer.get(), mAuxBuffer.get(), format);
 
            // Integrate signals
            Add(mSigBuffer.get(), mAuxBuffer.get(), cSiglen);
 
            // Do some dummy reads to break signal coherence
            mReader.ReadSamples(mRawBuffer.get(), n + 1, stride, format.format, format.endian);
         }
      }
 
      n++;
 
   } while ((n < cNumInts) && (actRead == cSiglen));
 
}

References Add(), cNumInts, ConvertSamples(), cSiglen, anonymous_namespace{ExportPCM.cpp}::format, MachineEndianness::Little, mAuxBuffer, mRawBuffer, mReader, mSigBuffer, MultiFormatReader::ReadSamples(), MultiFormatReader::Reset(), and MultiFormatReader::Uint8.

Referenced by Run().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ Run()

void FormatClassifier::Run ( )

private

Definition at line 129 of file FormatClassifier.cpp.

{
   // Calc the mono feature vector
   for (unsigned int n = 0; n < mClasses.size(); n++)
   {
      // Read the signal
      ReadSignal(mClasses[n], 1);
#ifdef FORMATCLASSIFIER_SIGNAL_DEBUG
      mpWriter->WriteSignal(mSigBuffer, cSiglen);
#endif
 
      // Do some simple preprocessing
      // Remove DC offset
      float smean = Mean(mSigBuffer.get(), cSiglen);
      Sub(mSigBuffer.get(), smean, cSiglen);
      // Normalize to +- 1.0
      Abs(mSigBuffer.get(), mAuxBuffer.get(), cSiglen);
      float smax = Max(mAuxBuffer.get(), cSiglen);
      Div(mSigBuffer.get(), smax, cSiglen);
 
      // Now actually fill the feature vector
      // Low to high band power ratio
      float pLo = mMeter.CalcPower(mSigBuffer.get(), 0.15f, 0.3f);
      float pHi = mMeter.CalcPower(mSigBuffer.get(), 0.45f, 0.1f);
      mMonoFeat[n] = pLo / pHi;
   }
 
   // Calc the stereo feature vector
   for (unsigned int n = 0; n < mClasses.size(); n++)
   {
      // Read the signal
      ReadSignal(mClasses[n], 2);
#ifdef FORMATCLASSIFIER_SIGNAL_DEBUG
      mpWriter->WriteSignal(mSigBuffer, cSiglen);
#endif
 
      // Do some simple preprocessing
      // Remove DC offset
      float smean = Mean(mSigBuffer.get(), cSiglen);
      Sub(mSigBuffer.get(), smean, cSiglen);
      // Normalize to +- 1.0
      Abs(mSigBuffer.get(), mAuxBuffer.get(), cSiglen);
      float smax = Max(mAuxBuffer.get(), cSiglen);
      Div(mSigBuffer.get(), smax, cSiglen);
 
      // Now actually fill the feature vector
      // Low to high band power ratio
      float pLo = mMeter.CalcPower(mSigBuffer.get(), 0.15f, 0.3f);
      float pHi = mMeter.CalcPower(mSigBuffer.get(), 0.45f, 0.1f);
      mStereoFeat[n] = pLo / pHi;
   }
 
   // Get the results
   size_t midx, sidx;
   float monoMax = Max(mMonoFeat.get(), mClasses.size(), &midx);
   float stereoMax = Max(mStereoFeat.get(), mClasses.size(), &sidx);
 
   if (monoMax > stereoMax)
   {
      mResultChannels = 1;
      mResultFormat = mClasses[midx];
   }
   else
   {
      mResultChannels = 2;
      mResultFormat = mClasses[sidx];
   }
 
}

References Abs(), SpecPowerCalculation::CalcPower(), cSiglen, Div(), mAuxBuffer, Max(), mClasses, Mean(), mMeter, mMonoFeat, mResultChannels, mResultFormat, mSigBuffer, mStereoFeat, ReadSignal(), and Sub().

Referenced by FormatClassifier().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ Sub()

void FormatClassifier::Sub	(	float *	in,
		float	subt,
		size_t	len
	)

private

Definition at line 276 of file FormatClassifier.cpp.

{
   for (unsigned int n = 0; n < len; n++)
   {
      in[n] -= subt;
   }
}

Referenced by Run().

Here is the caller graph for this function:

◆ ToFloat()

template<class T >

void FormatClassifier::ToFloat	(	T *	in,
		float *	out,
		size_t	len
	)

private

Definition at line 345 of file FormatClassifier.cpp.

{
   for(unsigned int n = 0; n < len; n++)
   {
      out[n] = (float) in[n];
   }
}

Referenced by ConvertSamples().

Here is the caller graph for this function:

Member Data Documentation

◆ cNumInts

const size_t FormatClassifier::cNumInts = 32

staticprivate

Definition at line 72 of file FormatClassifier.h.

Referenced by ReadSignal().

◆ cSiglen

const size_t FormatClassifier::cSiglen = 512

staticprivate

Definition at line 71 of file FormatClassifier.h.

Referenced by ConvertSamples(), ReadSignal(), and Run().

◆ mAuxBuffer

Floats FormatClassifier::mAuxBuffer { cSiglen }

private

Definition at line 83 of file FormatClassifier.h.

Referenced by ReadSignal(), and Run().

◆ mClasses

FormatVectorT FormatClassifier::mClasses

private

Definition at line 74 of file FormatClassifier.h.

Referenced by FormatClassifier(), and Run().

◆ mMeter

SpecPowerCalculation FormatClassifier::mMeter

private

Definition at line 76 of file FormatClassifier.h.

Referenced by Run().

◆ mMonoFeat

Floats FormatClassifier::mMonoFeat

private

Definition at line 86 of file FormatClassifier.h.

Referenced by FormatClassifier(), and Run().

◆ mRawBuffer

ArrayOf<uint8_t> FormatClassifier::mRawBuffer { cSiglen * 8 }

private

Definition at line 84 of file FormatClassifier.h.

Referenced by ReadSignal().

◆ mReader

MultiFormatReader FormatClassifier::mReader

private

Definition at line 75 of file FormatClassifier.h.

Referenced by ReadSignal().

◆ mResultChannels

unsigned FormatClassifier::mResultChannels { 0 }

private

Definition at line 90 of file FormatClassifier.h.

Referenced by GetResultChannels(), and Run().

◆ mResultFormat

FormatClassT FormatClassifier::mResultFormat

private

Definition at line 89 of file FormatClassifier.h.

Referenced by GetResultFormat(), GetResultFormatLibSndfile(), and Run().

◆ mSigBuffer

Floats FormatClassifier::mSigBuffer { cSiglen }

private

Definition at line 82 of file FormatClassifier.h.

Referenced by ReadSignal(), and Run().

◆ mStereoFeat

Floats FormatClassifier::mStereoFeat

private

Definition at line 87 of file FormatClassifier.h.

Referenced by FormatClassifier(), and Run().

The documentation for this class was generated from the following files:

Classes

Public Types

Public Member Functions

Private Member Functions

Private Attributes

Static Private Attributes

Detailed Description

Member Typedef Documentation

◆ FormatVectorIt

◆ FormatVectorT

Constructor & Destructor Documentation

◆ FormatClassifier()

◆ ~FormatClassifier()

Member Function Documentation

◆ Abs()

◆ Add()

◆ ConvertSamples()

◆ Div()

◆ GetResultChannels()

◆ GetResultFormat()

◆ GetResultFormatLibSndfile()

◆ Max() [1/2]

◆ Max() [2/2]

◆ Mean()

◆ ReadSignal()

◆ Run()

◆ Sub()

◆ ToFloat()

Member Data Documentation

◆ cNumInts

◆ cSiglen

◆ mAuxBuffer

◆ mClasses

◆ mMeter

◆ mMonoFeat

◆ mRawBuffer

◆ mReader

◆ mResultChannels

◆ mResultFormat

◆ mSigBuffer

◆ mStereoFeat