libsent/src/adin/adin_file.c

Go to the documentation of this file.

/*
 * Copyright (c) 1991-2007 Kawahara Lab., Kyoto University
 * Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and Technology
 * Copyright (c) 2005-2007 Julius project team, Nagoya Institute of Technology
 * All rights reserved
 */


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

static FILE *gfp;               
static boolean wav_p;           
static int maxlen;              
static int nowlen;              

static SP16 pre_data[2];
static boolean has_pre;         

static unsigned int sfreq;      

static char speechfilename[MAXPATHLEN]; 


/* read .wav data with endian conversion */
/* (all .wav datas are in little endian) */

static boolean
myread(void *buf, size_t unitbyte, int unitnum, FILE *fp)
{
  int tmp;
  if ((tmp = myfread(buf, unitbyte, unitnum, fp)) < unitnum) {
    return(FALSE);
  }
#ifdef WORDS_BIGENDIAN
  swap_bytes(buf, unitbyte, unitnum);
#endif
  return(TRUE);
}
#define MYREAD(A,B,C,D)  if (!myread(A, B, C, D)) {jlog("Error: adin_file: file is corrupted\n"); return -1;}

static boolean
setup_wav(FILE *fp)
{
  char dummy[9];
  unsigned int i, len;
  unsigned short s;

  /* 4 byte: byte num of rest ( = filesize - 8) */
  /* --- just skip them */
  MYREAD(dummy, 1, 4, fp);
  /* first part: WAVE format specifications */
  /* 4 byte: "WAVE" */
  MYREAD(dummy, 1, 4, fp);
  if (dummy[0] != 'W' ||
      dummy[1] != 'A' ||
      dummy[2] != 'V' ||
      dummy[3] != 'E') {
    jlog("Error: adin_file: WAVE header not found, file corrupted?\n");
    fclose_readfile(fp); return FALSE;
  }
  /* format chunk: "fmt " */
  MYREAD(dummy, 1, 4, fp);
  if (dummy[0] != 'f' ||
      dummy[1] != 'm' ||
      dummy[2] != 't' ||
      dummy[3] != ' ') {
    jlog("Error: adin_file: fmt chunk not found, file corrupted?\n");
    fclose_readfile(fp); return FALSE;
  }
  /* 4byte: byte size of this part */
  MYREAD(&len, 4, 1, fp);

  /* 2byte: data format */
  MYREAD(&s, 2, 1, fp);
  if (s != 1) {
    jlog("Error: adin_file: data format != PCM (id=%d)\n", s);
    fclose_readfile(fp); return FALSE;
  }
  /* 2byte: channel num */
  MYREAD(&s, 2, 1, fp);
  if (s >= 2) {
    jlog("Error: adin_file: channel num != 1 (%d)\n", s);
    fclose_readfile(fp); return FALSE;
  }
  /* 4byte: sampling rate */
  MYREAD(&i, 4, 1, fp);
  if (i != sfreq) {
    jlog("Error: adin_file: sampling rate != %d (%d)\n", sfreq, i);
    fclose_readfile(fp); return FALSE;
  }
  /* 4byte: bytes per second */
  MYREAD(&i, 4, 1, fp);
  if (i != sfreq * sizeof(SP16)) {
    jlog("Error: adin_file: bytes per second != %d (%d)\n", sfreq * sizeof(SP16), i);
    fclose_readfile(fp); return FALSE;
  }
  /* 2bytes: bytes per frame ( = (bytes per sample) x channel ) */
  MYREAD(&s, 2, 1, fp);
  if (s != 2) {
    jlog("Error: adin_file: (bytes per sample) x channel != 2 (%d)\n", s);
    fclose_readfile(fp); return FALSE;
  }
  /* 2bytes: bits per sample */
  MYREAD(&s, 2, 1, fp);
  if (s != 16) {
    jlog("Error: adin_file: bits per sample != 16 (%d)\n", s);
    fclose_readfile(fp); return FALSE;
  }
  /* skip rest */
  if (len > 16) {
    len -= 16;
    while (len > 0) {
      if (len > 8) {
        MYREAD(dummy, 1, 8, fp);
        len -= 8;
      } else {
        MYREAD(dummy, 1, len, fp);
        len = 0;
      }
    }
  }
  /* end of fmt part */

  /* seek for 'data' part */
  while (myread(dummy, 1, 4, fp)) {
    MYREAD(&len, 4, 1, fp);
    if (dummy[0] == 'd' &&
        dummy[1] == 'a' &&
        dummy[2] == 't' &&
        dummy[3] == 'a') {
      break;
    }
    for (i=0;i<len;i++) myread(dummy, 1, 1, fp);
  }
  /* ready to read in "data" part --- this is speech data */
  maxlen = len / sizeof(SP16);
  nowlen = 0;
  return TRUE;
}

static boolean
adin_file_open(char *filename)  /* NULL for standard input */
{
  FILE *fp;
  char dummy[4];

  if (filename != NULL) {
    if ((fp = fopen_readfile(filename)) == NULL) {
      jlog("Error: adin_file: failed to open %s\n",filename);
      return(FALSE);
    }
  } else {
    fp = stdin;
#if defined(_WIN32) && !defined(__CYGWIN32__) && !defined(__MINGW32__)
    if (_setmode( _fileno( stdin ), _O_BINARY ) == -1) {
      jlog("Error: adin_file: _setmode() failed\n");
    }
#endif
  }
    
  /* check first 4 byte to detect Microsoft WAVE format */
  if (myfread(dummy, 1, 4, fp) < 4) {
    jlog("Error: adin_file: size less than 4 bytes?\n",filename);
    fclose_readfile(fp);
    return(FALSE);
  }
  if (dummy[0] == 'R' &&
      dummy[1] == 'I' &&
      dummy[2] == 'F' &&
      dummy[3] == 'F') {
    /* it's a WAVE file */
    wav_p = TRUE;
    has_pre = FALSE;
    if (setup_wav(fp) == FALSE) {
      jlog("Error: adin_file: failed to read %s as a wav file\n",filename);
      fclose_readfile(fp);
      return(FALSE);
    }
  } else {
    /* read as raw format file */
    wav_p = FALSE;
    memcpy(pre_data, dummy, 4);    /* already read (4/sizeof(SP)) samples */
    has_pre = TRUE;
  }

  gfp = fp;

  return(TRUE);
}

static boolean
adin_file_close()
{
  FILE *fp;

  fp = gfp;
  if (fclose_readfile(fp) != 0) {
    jlog("Error: adin_file: failed to close file\n");
    return FALSE;
  }
 return TRUE; 
}

static boolean from_file;       
static FILE *fp_list;           

boolean
adin_file_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_file: 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);
}

boolean
adin_file_begin()
{
  boolean readp;

  /* 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 */
      }
    }
    /* open input file */
    if (adin_file_open(speechfilename) == FALSE) {
      jlog("Error: adin_file: failed to read speech data: \"%s\"\n", speechfilename);
    } else {
      jlog("Stat: adin_file: input speechfile: %s\n", speechfilename);
      readp = TRUE;
    }
  }
  return TRUE;
}

int
adin_file_read(SP16 *buf, int sampnum)
{
  FILE *fp;
  int cnt;

  fp = gfp;
  
  if (wav_p) {
    cnt = myfread(buf, sizeof(SP16), sampnum, fp);
    if (nowlen + cnt > maxlen) {
      cnt = maxlen - nowlen;
    }
    nowlen += cnt;
  } else {
    if (has_pre) {
      buf[0] = pre_data[0]; buf[1] = pre_data[1];
      has_pre = FALSE;
      cnt = myfread(&(buf[2]), sizeof(SP16), sampnum - 2, fp);
      if (cnt > 0) cnt += 2;
    } else {
      cnt = myfread(buf, sizeof(SP16), sampnum, fp);
    }
  }
  if (cnt == 0) {               /* error or EOF */
    if (myfeof(fp) == 1) {              /* EOF */
      return -1;
    }
    jlog("Error: adin_file: an error occured file reading file\n");
    adin_file_close();
    return -2;          /* error */
  }
  /* all .wav data are in little endian */
  /* assume .raw data are in big endian */
#ifdef WORDS_BIGENDIAN
  if (wav_p) swap_sample_bytes(buf, cnt);
#else
  if (!wav_p) swap_sample_bytes(buf, cnt);
#endif
  return cnt;
}

boolean
adin_file_end()
{
  /* nothing needed */
  adin_file_close();
  return TRUE;
}

boolean
adin_stdin_standby(int freq, void *arg)
{
  /* store sampling frequency */
  sfreq = freq;
  return(TRUE);
}

boolean
adin_stdin_begin()
{
  if (feof(stdin)) {            /* already reached the end of input stream */
    return FALSE;               /* terminate search here */
  } else {
    /* open input stream */
    if (adin_file_open(NULL) == FALSE) {
      jlog("Error: adin_stdin: failed to read speech data from stdin\n");
      return FALSE;
    }
    jlog("Stat: adin_stding: reading wavedata from stdin...\n");
  }
  return TRUE;
}

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

  if (wav_p) {
    cnt = myfread(buf, sizeof(SP16), sampnum, stdin);
  } else {
    if (has_pre) {
      buf[0] = pre_data[0]; buf[1] = pre_data[1];
      has_pre = FALSE;
      cnt = fread(&(buf[2]), sizeof(SP16), sampnum - 2, stdin);
      if (cnt > 0) cnt += 2;
    } else {
      cnt = fread(buf, sizeof(SP16), sampnum, stdin);
    }
  }
  if (cnt == 0) {     
    if (ferror(stdin)) {                /* error */
      jlog("Error: adin_file: an error occured file reading file\n");
      return -2;                /* error */
    }
    return -1;                  /* EOF */
  }
  /* all .wav data are in little endian */
  /* assume .raw data are in big endian */
#ifdef WORDS_BIGENDIAN
  if (wav_p) swap_sample_bytes(buf, cnt);
#else
  if (!wav_p) swap_sample_bytes(buf, cnt);
#endif
  return cnt;
}

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

---