www.pudn.com > ETSI_ES_202_212_software.rar > ExtAdvFrontEnd.c
/*===============================================================================
* ETSI ES 202 212 Distributed Speech Recognition
* Extended Advanced Front-End Feature Extraction Algorithm & Compression Algorithm
* Speech Reconstruction Algorithm.
* C-language software implementation
* Version 1.1.1 October, 2003
*===============================================================================*/
/*---------------------------------------------------------------------------
*
* FILE NAME: ExtAdvFrontEnd.c
* PURPOSE: This file contains the main part of Extended Advanced DSR FE.
* Speech samples are read from input waveform file frame by frame.
* Feature extraction is performed for each frame by calling the
* function DoAdvProcess () (see ParmInterface.c).
* Feature vectors are output to file in HTK format. VAD information
* is output to file in ASCII format.
* Command line arguments are handled by a command line parsing
* function.
*
*---------------------------------------------------------------------------*/
/*-----------------
* File Inclusions
*-----------------*/
#include
#include
#include
#include
#include "fileio.h"
#include "ParmInterface.h"
#include "pitchInterface.h"
#define VERSION_STR "Version 1.1.1, October 2003"
/*-------------------------------
* Global Definitions and Macros
*-------------------------------*/
#define PRINTMOD 15
#define IEEE_LE 0
#define IEEE_BE 1
/*---------------
* HTK constants
*---------------*/
#define HTK_PR_MFCC 6
#define HTK_PR_E 0x40
#define HTK_PR_0 0x2000
/*----------------------------------------------
* Global Variable Definitions and Declarations
*----------------------------------------------*/
BOOLEAN QuietMode = FALSE, /* Supress output to stderr */
FsSpecified = FALSE, /* Sampling frequency specified */
SwapSpecified = FALSE, /* Byte swap for raw data files specified */
InputKindSpecified = FALSE, /* Input file format specified */
NoOutHeaderSpecified = FALSE, /* No output HTK header option specified */
Noc0 = FALSE, /* No c0 coefficient to output feature vector */
NologE = FALSE; /* No logE component to output feature vector */
FILE *fp_in = NULL, /* Input HTK, NIST or raw data file */
*fp_out_htk = NULL, /* Output HTK file */
*fp_out_vad = NULL, /* Output VAD file */
*fp_out_pitch = NULL, /* Output pitch file */
*fp_out_class = NULL; /* Output class file */
int HeaderBytesToSkip = 0;
char InFilename[199], /* Name of input file */
OutFilename[199], /* Name of output file */
VADFilename[199], /* Name of output VAD file */
InputKind[10] = "NIST"; /* Input file format */
int SamplingFrequency = 8000, /* SamplingFrequency */
NativeByteOrder = IEEE_LE, /* Native byte ordering */
InputByteOrder, /* Default input byte ordering */
OutParmKind=HTK_PR_MFCC + HTK_PR_E + HTK_PR_0; /* Output parameter kind MFCC_0_E as default */
/*------------
* Prototypes
*------------*/
static int ReadBufWave (FILE *fp_in, FILE_TYPE *buf, int nSamples, int Swap);
static BOOLEAN ParseCommLine (int argc, char *argv[]);
/*-----------
* Functions
*-----------*/
/*----------------------------------------------------------------------------
* FUNCTION NAME: ReadBufWave
*
* PURPOSE: Read buffer from input file
*
* INPUT:
* fp_in Pointer to file
* nSamples Number of samples to read
* swap Byte swap specification
*
* OUTPUT
* buf Pointer to buffer
*
* RETURN VALUE
* FALSE In case of fread() error
* TRUE Otherwise
*
*---------------------------------------------------------------------------*/
static int ReadBufWave (FILE *fp_in, FILE_TYPE *buf, int nSamples, int Swap)
{
FILE_TYPE s;
int i;
for (i=0; i> 8);
buf[i] = (FILE_TYPE)s;
}
return TRUE;
}
/*----------------------------------------------------------------------------
* FUNCTION NAME: ParseCommLine
*
* PURPOSE: Parses command line arguments, opens input and output files
*
* INPUT:
* argc Number of command line arguments
* argv Array of command line arguments
*
* OUTPUT
* none
*
* RETURN VALUE
* FALSE In case of any errors
* TRUE Otherwise
*
*---------------------------------------------------------------------------*/
static BOOLEAN ParseCommLine (int argc, char *argv[])
{
int mark = 0, mand_arg_no = 1;
for ( ; argc; argc--, mark++)
{
if (strcmp (argv[mark], "-q") == 0)
{
QuietMode = TRUE;
}
else if (strcmp (argv[mark], "-fs") == 0)
{
FsSpecified = TRUE;
--argc;
++mark;
SamplingFrequency = 1000 * atoi (argv[mark]);
}
else if (strcmp (argv[mark], "-swap") == 0)
{
SwapSpecified = TRUE;
}
else if (strcmp (argv[mark], "-F") == 0)
{
InputKindSpecified = TRUE;
--argc;
++mark;
strcpy (InputKind, argv[mark]);
}
else if (strcmp (argv[mark], "-noh") == 0)
{
NoOutHeaderSpecified = TRUE;
}
else if (strcmp (argv[mark], "-noc0") == 0)
{
Noc0 = TRUE;
}
else if (strcmp (argv[mark], "-nologE") == 0)
{
NologE = TRUE;
}
else if (strcmp (argv[mark], "-skip_header_bytes") == 0)
{
--argc;
++mark;
HeaderBytesToSkip = atoi (argv[mark]);
}
else if (argv[mark][0] == '-')
{
fprintf (stderr, "WARNING: Un-recognized flag '%s' !\r\n", argv[mark]);
}
else
{
switch (mand_arg_no)
{
case 1: /* First mandatory argument - input file */
strcpy (InFilename, argv[mark]);
fp_in = fopen (argv[mark], "rb");
if (fp_in == NULL)
{
fprintf (stderr, " ERROR: Could not open file '%s' !\r\n",
argv[mark]);
return FALSE;
}
break;
case 2: /* Second mandatory argument - output HTK file */
strcpy (OutFilename, argv[mark]);
fp_out_htk = fopen (OutFilename, "wb");
if (fp_out_htk == NULL)
{
fprintf (stderr, " ERROR: Could not open file '%s' !\r\n",
OutFilename);
return FALSE;
}
strcpy (VADFilename, strcat(strtok(argv[mark], "."), ".vad"));
fp_out_vad = fopen (VADFilename, "wb");
if (fp_out_vad == NULL)
{
fprintf (stderr, "ERROR: Could not open file '%s' !\r\n", VADFilename);
return FALSE;
}
break;
case 3: /* Third mandatory argument - output pitch file */
strcpy (OutFilename, argv[mark]);
fp_out_pitch = fopen (OutFilename, "wb");
if (fp_out_pitch == NULL)
{
fprintf (stderr, " ERROR: Could not open file '%s' !\r\n",
OutFilename);
return FALSE;
}
break;
case 4: /* Fourth mandatory argument - output class file */
strcpy (OutFilename, argv[mark]);
fp_out_class = fopen (OutFilename, "w"); // ASCII output
if (fp_out_class == NULL)
{
fprintf (stderr, " ERROR: Could not open file '%s' !\r\n",
OutFilename);
return FALSE;
}
break;
default:
fprintf (stderr, " ERROR: Too many input arguments!\r\n");
return FALSE;
} /* end switch on mandatory argument no */
mand_arg_no++;
} // end of loop on arguments
}
if (!fp_in || !fp_out_htk || !fp_out_pitch || !fp_out_class)
{
fprintf (stderr, "ERROR: Input and output files must be given!\r\n");
return FALSE;
}
if (strcmp (InputKind, "NIST") && strcmp (InputKind, "HTK") && strcmp (InputKind, "RAW"))
{
fprintf (stderr, "ERROR: Invalid input file format '%s'!\r\n", InputKind);
return FALSE;
}
if (strcmp (InputKind, "RAW") && HeaderBytesToSkip != 0)
{
fprintf(stderr, "ERROR: skip_header_bytes can be specified for RAW format only.\r\n");
return FALSE;
}
if (strcmp (InputKind, "RAW") && FsSpecified)
fprintf (stderr, "WARNING: Sampling frequency needs to be specified only for raw data files.\r\n");
if (strcmp (InputKind, "RAW") && SwapSpecified)
fprintf (stderr, "WARNING: Byte swapping needs to be specified only for raw data files if necessary.\r\n");
if (Noc0 && NologE) TextToParmKind( "MFCC", &OutParmKind);
else if (Noc0) TextToParmKind( "MFCC_E", &OutParmKind);
else if (NologE) TextToParmKind( "MFCC_0", &OutParmKind);
return TRUE;
}
/*----------------------------------------------------------------------------
* FUNCTION NAME: main
*
* PURPOSE: Main front-end operations from input speech samples to output
* feature vectors. See embedded comments.
*
* INPUT:
* argc Number of command line arguments (passed to ParseCommLine)
* argv Array of command line arguments (passed to ParseCommLine)
*
* OUTPUT
* none
*
* RETURN VALUE
* TRUE In case of any errors
* FALSE Otherwise
*
*---------------------------------------------------------------------------*/
extern int main (int argc, char *argv[])
{
FILE_TYPE *SigBuf;
X_FLOAT32 FeatureBuffer[NUM_CEP_COEFF + 2];
FEParamsX *pFEParX;
HTK_Header InHheader, OutHheader;
NIST_Header InNheader;
int rc;
int NbSamplesToRead;
long FrameCounter = 0;
long SpeechFrameCounter = 0;
long NonSpeechFrameCounter = 0;
if (!ParseCommLine (argc - 1, argv + 1))
{
fprintf (stderr,"\r\n ETSI ES 202 212 DSR Extended Advanced Front-End Feature Extraction Algorithm");
fprintf (stderr,"\r\n C-language software implementation");
fprintf (stderr,"\r\n %s\r\n\n",VERSION_STR);
fprintf (stderr, "\r\n USAGE:");
fprintf (stderr, " %s infile HTK_outfile pitch_outfile class_outfile [options]\r\n", argv[0]);
fprintf (stderr, "\n The program outputs also .vad ASCII VAD file\r\n");
fprintf (stderr, "\r\n OPTIONS:\r\n");
fprintf (stderr, " -q Quiet Mode (%s)\r\n", QuietMode ? "TRUE" : "FALSE");
fprintf (stderr, " -F format Input file format (NIST,HTK,RAW) (%s)\r\n", InputKind);
fprintf (stderr, " -fs freq Sampling frequency in kHz (%d,%d,%d) (%d)\r\n", SAMPLING_FREQ_1, SAMPLING_FREQ_2, SAMPLING_FREQ_3, SamplingFrequency / 1000);
fprintf (stderr, " -swap Change input byte ordering (%s)\r\n", SwapSpecified ? "Swapped" : "Native");
fprintf (stderr, " (Native byte ordering is %s)\r\n", NativeByteOrder ? "ieee-be" : "ieee-le");
fprintf (stderr, " -noh No HTK header to output file (%s)\r\n", NoOutHeaderSpecified ? "TRUE" : "FALSE" );
fprintf (stderr, " -noc0 No c0 coefficient to output feature vector (%s)\r\n", Noc0 ? "TRUE" : "FALSE" );
fprintf (stderr, " -nologE No logE component to output feature vector (%s)\r\n", NologE ? "TRUE" : "FALSE" );
fprintf (stderr, " -skip_header_bytes n - Skip header, first n bytes (Only for F RAW)\r\n");
return TRUE;
}
if (!QuietMode)
{
fprintf (stderr,"\r\n ETSI ES 202 212 DSR Extended Advanced Front-End Feature Extraction Algorithm");
fprintf (stderr,"\r\n C-language software implementation");
fprintf (stderr,"\r\n %s\r\n\n",VERSION_STR);
}
/*----------------*/
/* Initialization */
/*----------------*/
InputByteOrder = NativeByteOrder;
/*-----------------------------------------------------------------
* Read input header, extract sampling frequency and byte ordering
*-----------------------------------------------------------------*/
if (!strcmp (InputKind, "NIST"))
{
if (!ReadNISTHeader (fp_in, &InNheader))
{
fprintf (stderr, "ERROR: Invalid NIST header !\r\n");
return TRUE;
}
SamplingFrequency = InNheader.SampleRate;
if (strcmp (InNheader.SampleByteFormat, "10"))
InputByteOrder = IEEE_LE;
else
InputByteOrder = IEEE_BE;
}
else if (!strcmp (InputKind, "HTK"))
{
if (!ReadHTKHeader (fp_in, &InHheader, InputByteOrder!=IEEE_BE))
{
fprintf (stderr, "ERROR: Invalid HTK header !\r\n");
return TRUE;
}
// 625->16kHz, 1250->8kHz, 909->11kHz
SamplingFrequency = 10 * floor ((float) 1e6 / (float) InHheader.sampPeriod);
InputByteOrder = IEEE_BE;
}
/*------------------------------------------------------------
* Write output header (number of frames to be updated later)
*------------------------------------------------------------*/
OutHheader.nSamples = 0;
OutHheader.sampPeriod = 100000; /* 10000.0 us (100 Hz) */
OutHheader.sampSize = (NUM_CEP_COEFF - (Noc0 ? 1:0) + (NologE ? 0:1)) * 4;
OutHheader.sampKind = OutParmKind;
if (!NoOutHeaderSpecified) WriteHTKHeader (fp_out_htk, &OutHheader);
/*------------------------------------------
* Memory allocation for FE data structures
*------------------------------------------*/
pFEParX = AdvProcessAlloc (SamplingFrequency);
/* For pitch and class extraction */
pFEParX->pfInpSpeech = (X_FLOAT32*)malloc(sizeof(X_FLOAT32)*FRAME_LENGTH);
if ( !(pFEParX->pfInpSpeech))
{
fprintf (stderr, "ERROR: Memory allocation error occured!\r\n");
return TRUE;
}
pFEParX->pfUBSpeech = (X_FLOAT32*)malloc(sizeof(X_FLOAT32)*FRAME_LENGTH);
if ( !(pFEParX->pfUBSpeech))
{
fprintf (stderr, "ERROR: Memory allocation error occured!\r\n");
return TRUE;
}
pFEParX->pfProcSpeech = (X_FLOAT32*)malloc(sizeof(X_FLOAT32)*(FRAME_LENGTH+HISTORY_LENGTH));
if ( !(pFEParX->pfProcSpeech))
{
fprintf (stderr, "ERROR: Memory allocation error occured!\r\n");
return TRUE;
}
pFEParX->pfDownSampledProcSpeech = (X_FLOAT32*)
malloc(sizeof(X_FLOAT32)*(FRAME_LENGTH+HISTORY_LENGTH)/DOWN_SAMP_FACTOR+1);
if ( !(pFEParX->pfDownSampledProcSpeech))
{
fprintf (stderr, "ERROR: Memory allocation error occured!\r\n");
return TRUE;
}
/*-------------------------------------------------------
* Initialization of FE data structures and input buffer
*-------------------------------------------------------*/
AdvProcessInit (pFEParX);
rc = InitPitchRom(&(pFEParX->pPitchRom));
if (rc != 0)
{
fprintf(stderr,"ERROR: Can't initialize PITCH ROM, RC = %d\r\n",rc);
return TRUE;
}
rc = InitPitchEstimator(pFEParX->pPitchRom,&(pFEParX->pPitchEstimator));
if ( rc!= 0)
{
fprintf (stderr, "ERROR: can't initialize PITCH ESTIMATOR, RC = %d\r\n",rc);
return TRUE;
}
pFEParX->speech_fid = fp_in;
pFEParX->htk_fid = fp_out_htk;
pFEParX->vad_fid = fp_out_vad;
pFEParX->pitch_fid = fp_out_pitch;
pFEParX->class_fid = fp_out_class;
NbSamplesToRead = pFEParX->NbSamplesToRead;
SigBuf = calloc (1, sizeof (SigBuf[0]) * NbSamplesToRead);
if (SigBuf == NULL)
{
fprintf (stderr,"ERROR: Memory allocation error occured!\r\n");
return TRUE;
}
/*------------
* Processing
*------------*/
fseek(fp_in,HeaderBytesToSkip,SEEK_SET);
while (ReadBufWave (fp_in, SigBuf, NbSamplesToRead,
(SwapSpecified || InputByteOrder != NativeByteOrder)))
{
FrameCounter++;
if (DoAdvProcess (SigBuf, FeatureBuffer, pFEParX))
{
/*---------------
* Output result
*---------------*/
WriteHTKFeature (fp_out_htk, FeatureBuffer, (short) (NUM_CEP_COEFF - (Noc0 ? 1:0) + (NologE ? 0:1)));
fprintf (fp_out_vad, "%d ", pFEParX->VAD);
if (pFEParX->VAD == SPEECH_FRAME)
SpeechFrameCounter++;
else
NonSpeechFrameCounter++;
}
/*---------------------------
* Display processing status
*---------------------------*/
if (!QuietMode && !(FrameCounter % PRINTMOD))
{
fprintf (stderr, "\rProcessing status: %ld frames ...", FrameCounter);
fflush (stderr);
}
}
/*----------
* Flushing
*----------*/
while (FlushAdvProcess (FeatureBuffer, pFEParX))
{
/*---------------
* Output result
*---------------*/
WriteHTKFeature (fp_out_htk, FeatureBuffer, (short) (NUM_CEP_COEFF - (Noc0 ? 1:0) + (NologE ? 0:1)));
fprintf (fp_out_vad, "%d ", pFEParX->VAD);
if (pFEParX->VAD == SPEECH_FRAME)
SpeechFrameCounter++;
else
NonSpeechFrameCounter++;
}
/*----------------
* Memory release
*----------------*/
DeallocatePitchRom(pFEParX->pPitchRom);
DeallocatePitchEstimator(pFEParX->pPitchEstimator);
free(pFEParX->pfUBSpeech);
free(pFEParX->pfProcSpeech);
free(pFEParX->pfDownSampledProcSpeech);
AdvProcessDelete (&pFEParX);
free (SigBuf);
/*----------------------------------------
* Correct number of frames in HTK header
*----------------------------------------*/
OutHheader.nSamples = SpeechFrameCounter + NonSpeechFrameCounter;
if (!NoOutHeaderSpecified) WriteHTKHeader (fp_out_htk, &OutHheader);
/*------------------------------
* Close input and output files
*------------------------------*/
fclose (fp_in);
fclose (fp_out_htk);
fclose (fp_out_vad);
fclose (fp_out_pitch);
fclose (fp_out_class);
/*----------------------
* Display final status
*----------------------*/
if (!QuietMode)
fprintf (stderr, "\rProcessed: %ld Frames. \r\n", FrameCounter);
fprintf (stderr, "Extended Advanced Front End: ");
fprintf (stderr, InFilename);
fprintf (stderr, ": %ld Frames, %ld Speech, %ld Non Speech, %ld Lost.\r\n", FrameCounter,
SpeechFrameCounter, NonSpeechFrameCounter, FrameCounter - SpeechFrameCounter - NonSpeechFrameCounter);
if (SpeechFrameCounter == 0)
fprintf (stderr, "NO SPEECH DETECTED !\r\n");
return FALSE;
}