libsent/src/adin/adin_sndfile.c

説明を見る。

/*
 * Copyright (c) 1991-2011 Kawahara Lab., Kyoto University
 * Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and Technology
 * Copyright (c) 2005-2011 Julius project team, Nagoya Institute of Technology
 * All rights reserved
 */

#include <sent/stddefs.h>
#include <sent/speech.h>
#include <sent/adin.h>

#ifdef HAVE_LIBSNDFILE

/* sound header */
#include <sndfile.h>

static int sfreq;               
static SF_INFO sinfo;           
static SNDFILE *sp;             
static boolean from_file;       
static FILE *fp_list;           
static char speechfilename[MAXPATHLEN]; 

static boolean
check_format(SF_INFO *s)
{
  if ((s->format & SF_FORMAT_TYPEMASK) != SF_FORMAT_RAW) {
    if (s->samplerate != sfreq) {
      jlog("Error: adin_sndfile: sample rate != %d, it's %d Hz data\n", sfreq, s->samplerate);
      return FALSE;
    }
  }
  if (s->channels != 1) {
    jlog("Error: adin_sndfile: channel num != 1, it has %d channels\n", s->channels);
    return FALSE;
  }
#ifdef HAVE_LIBSNDFILE_VER1
  if ((s->format & SF_FORMAT_SUBMASK) != SF_FORMAT_PCM_16) {
    jlog("Error: adin_sndfile: not 16-bit data\n");
    return FALSE;
  }
#else
  if (s->pcmbitwidth != 16) {
    jlog("Error: adin_sndfile: not 16-bit data, it's %d bit\n", s->pcmbitwidth);
    return FALSE;
  }
#endif
  return TRUE;
}

static void
print_format(SF_INFO *s)
{
  switch(s->format & SF_FORMAT_TYPEMASK) {
  case SF_FORMAT_WAV:    jlog("Stat: adin_sndfile: input format = Microsoft WAV\n"); break;
  case SF_FORMAT_AIFF:   jlog("Stat: adin_sndfile: input format = Apple/SGI AIFF\n"); break;
  case SF_FORMAT_AU:     jlog("Stat: adin_sndfile: input format = Sun/NeXT AU\n"); break;
#ifndef HAVE_LIBSNDFILE_VER1
  case SF_FORMAT_AULE:   jlog("Stat: adin_sndfile: input format = DEC AU\n"); break;
#endif
  case SF_FORMAT_RAW:    jlog("Stat: adin_sndfile: input format = RAW\n"); break;
  case SF_FORMAT_PAF:    jlog("Stat: adin_sndfile: input format = Ensoniq PARIS\n"); break;
  case SF_FORMAT_SVX:    jlog("Stat: adin_sndfile: input format = Amiga IFF / SVX8 / SV16\n"); break;
  case SF_FORMAT_NIST:   jlog("Stat: adin_sndfile: input format = Sphere NIST\n"); break;
#ifdef HAVE_LIBSNDFILE_VER1
  case SF_FORMAT_VOC:    jlog("Stat: adin_sndfile: input format = VOC file\n"); break;
  case SF_FORMAT_IRCAM:  jlog("Stat: adin_sndfile: input format = Berkeley/IRCAM/CARL\n"); break;
  case SF_FORMAT_W64:    jlog("Stat: adin_sndfile: input format = Sonic Foundry's 64bit RIFF/WAV\n"); break;
  case SF_FORMAT_MAT4:   jlog("Stat: adin_sndfile: input format = Matlab (tm) V4.2 / GNU Octave 2.0\n"); break;
  case SF_FORMAT_MAT5:   jlog("Stat: adin_sndfile: input format = Matlab (tm) V5.0 / GNU Octave 2.1\n"); break;
#endif
  default: jlog("Stat: adin_sndfile: input format = UNKNOWN TYPE\n"); break;
  }
  switch(s->format & SF_FORMAT_SUBMASK) {
#ifdef HAVE_LIBSNDFILE_VER1
  case SF_FORMAT_PCM_U8:    jlog("Stat: adin_sndfile: input type = Unsigned 8 bit PCM\n"); break;
  case SF_FORMAT_PCM_S8:    jlog("Stat: adin_sndfile: input type = Signed 8 bit PCM\n"); break;
  case SF_FORMAT_PCM_16:    jlog("Stat: adin_sndfile: input type = Signed 16 bit PCM\n"); break;
  case SF_FORMAT_PCM_24:    jlog("Stat: adin_sndfile: input type = Signed 24 bit PCM\n"); break;
  case SF_FORMAT_PCM_32:    jlog("Stat: adin_sndfile: input type = Signed 32 bit PCM\n"); break;
  case SF_FORMAT_FLOAT:     jlog("Stat: adin_sndfile: input type = 32bit float\n"); break;
  case SF_FORMAT_DOUBLE:    jlog("Stat: adin_sndfile: input type = 64bit float\n"); break;
  case SF_FORMAT_ULAW:      jlog("Stat: adin_sndfile: input type = U-Law\n"); break;
  case SF_FORMAT_ALAW:      jlog("Stat: adin_sndfile: input type = A-Law\n"); break;
  case SF_FORMAT_IMA_ADPCM: jlog("Stat: adin_sndfile: input type = IMA ADPCM\n"); break;
  case SF_FORMAT_MS_ADPCM:  jlog("Stat: adin_sndfile: input type = Microsoft ADPCM\n"); break;
  case SF_FORMAT_GSM610:    jlog("Stat: adin_sndfile: input type = GSM 6.10, \n"); break;
  case SF_FORMAT_G721_32:   jlog("Stat: adin_sndfile: input type = 32kbs G721 ADPCM\n"); break;
  case SF_FORMAT_G723_24:   jlog("Stat: adin_sndfile: input type = 24kbs G723 ADPCM\n"); break;
  case SF_FORMAT_G723_40:   jlog("Stat: adin_sndfile: input type = 40kbs G723 ADPCM\n"); break;
#else
  case SF_FORMAT_PCM:       jlog("Stat: adin_sndfile: input type = PCM\n"); break;
  case SF_FORMAT_FLOAT:     jlog("Stat: adin_sndfile: input type = floats\n"); break;
  case SF_FORMAT_ULAW:      jlog("Stat: adin_sndfile: input type = U-Law\n"); break;
  case SF_FORMAT_ALAW:      jlog("Stat: adin_sndfile: input type = A-Law\n"); break;
  case SF_FORMAT_IMA_ADPCM: jlog("Stat: adin_sndfile: input type = IMA ADPCM\n"); break;
  case SF_FORMAT_MS_ADPCM:  jlog("Stat: adin_sndfile: input type = Microsoft ADPCM\n"); break;
  case SF_FORMAT_PCM_BE:    jlog("Stat: adin_sndfile: input type = Big endian PCM\n"); break;
  case SF_FORMAT_PCM_LE:    jlog("Stat: adin_sndfile: input type = Little endian PCM\n"); break;
  case SF_FORMAT_PCM_S8:    jlog("Stat: adin_sndfile: input type = Signed 8 bit PCM\n"); break;
  case SF_FORMAT_PCM_U8:    jlog("Stat: adin_sndfile: input type = Unsigned 8 bit PCM\n"); break;
  case SF_FORMAT_SVX_FIB:   jlog("Stat: adin_sndfile: input type = SVX Fibonacci Delta\n"); break;
  case SF_FORMAT_SVX_EXP:   jlog("Stat: adin_sndfile: input type = SVX Exponential Delta\n"); break;
  case SF_FORMAT_GSM610:    jlog("Stat: adin_sndfile: input type = GSM 6.10, \n"); break;
  case SF_FORMAT_G721_32:   jlog("Stat: adin_sndfile: input type = 32kbs G721 ADPCM\n"); break;
  case SF_FORMAT_G723_24:   jlog("Stat: adin_sndfile: input type = 24kbs G723 ADPCM\n"); break;
#endif
  default: jlog("Stat: adin_sndfile: input type = UNKNOWN SUBTYPE\n"); break;
  }

#ifdef HAVE_LIBSNDFILE_VER1
  switch(s->format & SF_FORMAT_ENDMASK) {
  case SF_ENDIAN_FILE:      jlog("Stat: adin_sndfile: endian = file native endian\n"); break;
  case SF_ENDIAN_LITTLE:    jlog("Stat: adin_sndfile: endian = forced little endian\n"); break;
  case SF_ENDIAN_BIG:       jlog("Stat: adin_sndfile: endian = forced big endian\n"); break;
  case SF_ENDIAN_CPU:       jlog("Stat: adin_sndfile: endian = forced CPU native endian\n"); break;
  }
  jlog("Stat: adin_sndfile: %d Hz, %d channels\n", s->samplerate, s->channels);
#else
  jlog("Stat: adin_sndfile: %d bit, %d Hz, %d channels\n", s->pcmbitwidth, s->samplerate, s->channels);
#endif
}

boolean
adin_sndfile_standby(int freq, void *arg)
{
  char *fname = arg;
  if (fname != NULL) {
    /* read input filename from file */
    if ((fp_list = fopen(fname, "r")) == NULL) {
      jlog("Error: adin_sndfile: failed to open %s\n", fname);
      return(FALSE);
    }
    from_file = TRUE;
  } else {
    /* read filename from stdin */
    from_file = FALSE;
  }
  /* store sampling frequency */
  sfreq = freq;
  
  return(TRUE);
}

static boolean
adin_sndfile_open(char *filename)
{
#ifndef HAVE_LIBSNDFILE_VER1
  sinfo.samplerate = sfreq;
  sinfo.pcmbitwidth = 16;
  sinfo.channels = 1;
#endif
  sinfo.format = 0x0;
  if ((sp = 
#ifdef HAVE_LIBSNDFILE_VER1
       sf_open(filename, SFM_READ, &sinfo)
#else
       sf_open_read(filename, &sinfo)
#endif
       ) == NULL) {
    /* retry assuming raw format */
    sinfo.samplerate = sfreq;
    sinfo.channels = 1;
#ifdef HAVE_LIBSNDFILE_VER1
    sinfo.format = SF_FORMAT_RAW | SF_FORMAT_PCM_16 | SF_ENDIAN_BIG;
#else
    sinfo.pcmbitwidth = 16;
    sinfo.format = SF_FORMAT_RAW | SF_FORMAT_PCM_BE;
#endif
    if ((sp =
#ifdef HAVE_LIBSNDFILE_VER1
         sf_open(filename, SFM_READ, &sinfo)
#else
         sf_open_read(filename, &sinfo)
#endif
         ) == NULL) {
      sf_perror(sp);
      jlog("Error: adin_sndfile: failed to open speech data: \"%s\"\n",filename);
    }
  }
  if (sp == NULL) {             /* open failure */
    return FALSE;
  }
  /* check its format */
  if (! check_format(&sinfo)) {
    return FALSE;
  }
  return TRUE;
}

boolean
adin_sndfile_begin(char *filename)
{
  boolean readp;

  if (filename != NULL) {
    if (adin_sndfile_open(filename) == FALSE) {
      jlog("Error: adin_sndfile: invalid format: \"%s\"\n", filename);
      print_format(&sinfo);
      return FALSE;
    }
    jlog("Stat: adin_sndfile: input speechfile: %s\n", filename);
    print_format(&sinfo);
    strcpy(speechfilename, filename);
    return TRUE;
  }

  /* ready to read next input */
  readp = FALSE;
  while(readp == FALSE) {
    if (from_file) {
      /* read file name from listfile */
      do {
        if (getl_fp(speechfilename, MAXPATHLEN, fp_list) == NULL) { /* end of input */
          fclose(fp_list);
          return(FALSE); /* end of input */
        }
      } while (speechfilename[0] == '#'); /* skip comment */
    } else {
      /* read file name from stdin */
      if (get_line_from_stdin(speechfilename, MAXPATHLEN, "enter filename->") == NULL) {
        return (FALSE); /* end of input */
      }
    }
    if (adin_sndfile_open(speechfilename) == FALSE) {
      jlog("Error: adin_sndfile: invalid format: \"%s\"\n",speechfilename);
      print_format(&sinfo);
    } else {
      jlog("Stat: adin_sndfile: input speechfile: %s\n",speechfilename);
      print_format(&sinfo);
      readp = TRUE;
    }
  }
  return TRUE;
}

int
adin_sndfile_read(SP16 *buf, int sampnum)
{
  int cnt;

  cnt = sf_read_short(sp, buf, sampnum);
  if (cnt == 0) {               /* EOF */
    return -1;
  } else if (cnt < 0) {         /* error */
    sf_perror(sp);
    sf_close(sp);
    return -2;          /* error */
  }
  return cnt;
}

boolean
adin_sndfile_end()
{
  /* close files */
  if (sf_close(sp) != 0) {
    sf_perror(sp);
    jlog("Error: adin_sndfile: failed to close\n");
    return FALSE;
  }
  return TRUE;
}

char *
adin_sndfile_get_current_filename()
{
  return(speechfilename);
}

#endif /* ~HAVE_LIBSNDFILE */