Exemple #1
0
/*
 * Initialize phones (ci and triphones) and state->senone mappings from .mdef file.
 */
mdef_t *
mdef_init(char *mdeffile, int32 breport)
{
    FILE *fp;
    int32 n_ci, n_tri, n_map, n;
    __BIGSTACKVARIABLE__ char tag[1024], buf[1024];
    uint16 **senmap;
    int p;
    int32 s, ci, cd;
    mdef_t *m;

    if (!mdeffile)
        E_FATAL("No mdef-file\n");

    if (breport)
        E_INFO("Reading model definition: %s\n", mdeffile);

    m = (mdef_t *) ckd_calloc(1, sizeof(mdef_t));       /* freed in mdef_free */

    if ((fp = fopen(mdeffile, "r")) == NULL)
        E_FATAL_SYSTEM("Failed to open mdef file '%s' for reading", mdeffile);

    if (noncomment_line(buf, sizeof(buf), fp) < 0)
        E_FATAL("Empty file: %s\n", mdeffile);

    if (strncmp(buf, "BMDF", 4) == 0 || strncmp(buf, "FDMB", 4) == 0) {
        E_INFO
            ("Found byte-order mark %.4s, assuming this is a binary mdef file\n",
             buf);
        fclose(fp);
        ckd_free(m);
        return NULL;
    }
    if (strncmp(buf, MODEL_DEF_VERSION, strlen(MODEL_DEF_VERSION)) != 0)
        E_FATAL("Version error: Expecing %s, but read %s\n",
                MODEL_DEF_VERSION, buf);

    /* Read #base phones, #triphones, #senone mappings defined in header */
    n_ci = -1;
    n_tri = -1;
    n_map = -1;
    m->n_ci_sen = -1;
    m->n_sen = -1;
    m->n_tmat = -1;
    do {
        if (noncomment_line(buf, sizeof(buf), fp) < 0)
            E_FATAL("Incomplete header\n");

        if ((sscanf(buf, "%d %s", &n, tag) != 2) || (n < 0))
            E_FATAL("Error in header: %s\n", buf);

        if (strcmp(tag, "n_base") == 0)
            n_ci = n;
        else if (strcmp(tag, "n_tri") == 0)
            n_tri = n;
        else if (strcmp(tag, "n_state_map") == 0)
            n_map = n;
        else if (strcmp(tag, "n_tied_ci_state") == 0)
            m->n_ci_sen = n;
        else if (strcmp(tag, "n_tied_state") == 0)
            m->n_sen = n;
        else if (strcmp(tag, "n_tied_tmat") == 0)
            m->n_tmat = n;
        else
            E_FATAL("Unknown header line: %s\n", buf);
    } while ((n_ci < 0) || (n_tri < 0) || (n_map < 0) ||
             (m->n_ci_sen < 0) || (m->n_sen < 0) || (m->n_tmat < 0));

    if ((n_ci == 0) || (m->n_ci_sen == 0) || (m->n_tmat == 0)
        || (m->n_ci_sen > m->n_sen))
        E_FATAL("%s: Error in header\n", mdeffile);

    /* Check typesize limits */
    if (n_ci >= MAX_INT16)
        E_FATAL("%s: #CI phones (%d) exceeds limit (%d)\n", mdeffile, n_ci,
                MAX_INT16);
    if (n_ci + n_tri >= MAX_INT32) /* Comparison is always false... */
        E_FATAL("%s: #Phones (%d) exceeds limit (%d)\n", mdeffile,
                n_ci + n_tri, MAX_INT32);
    if (m->n_sen >= MAX_INT16)
        E_FATAL("%s: #senones (%d) exceeds limit (%d)\n", mdeffile,
                m->n_sen, MAX_INT16);
    if (m->n_tmat >= MAX_INT32) /* Comparison is always false... */
        E_FATAL("%s: #tmats (%d) exceeds limit (%d)\n", mdeffile,
                m->n_tmat, MAX_INT32);

    m->n_emit_state = (n_map / (n_ci + n_tri)) - 1;
    if ((m->n_emit_state + 1) * (n_ci + n_tri) != n_map)
        E_FATAL
            ("Header error: n_state_map not a multiple of n_ci*n_tri\n");

    /* Initialize ciphone info */
    m->n_ciphone = n_ci;
    m->ciphone_ht = hash_table_new(n_ci, HASH_CASE_YES);  /* With case-insensitive string names *//* freed in mdef_free */
    m->ciphone = (ciphone_t *) ckd_calloc(n_ci, sizeof(ciphone_t));     /* freed in mdef_free */

    /* Initialize phones info (ciphones + triphones) */
    m->n_phone = n_ci + n_tri;
    m->phone = (phone_t *) ckd_calloc(m->n_phone, sizeof(phone_t));     /* freed in mdef_free */

    /* Allocate space for state->senone map for each phone */
    senmap = ckd_calloc_2d(m->n_phone, m->n_emit_state, sizeof(**senmap));      /* freed in mdef_free */
    m->sseq = senmap;           /* TEMPORARY; until it is compressed into just the unique ones */

    /* Allocate initial space for <ci,lc,rc,wpos> -> pid mapping */
    m->wpos_ci_lclist = (ph_lc_t ***) ckd_calloc_2d(N_WORD_POSN, m->n_ciphone, sizeof(ph_lc_t *));      /* freed in mdef_free */

    /*
     * Read base phones and triphones.  They'll simply be assigned a running sequence
     * number as their "phone-id".  If the phone-id < n_ci, it's a ciphone.
     */

    /* Read base phones */
    for (p = 0; p < n_ci; p++) {
        if (noncomment_line(buf, sizeof(buf), fp) < 0)
            E_FATAL("Premature EOF reading CIphone %d\n", p);
        parse_base_line(m, buf, p);
    }
    m->sil = mdef_ciphone_id(m, S3_SILENCE_CIPHONE);

    /* Read triphones, if any */
    for (; p < m->n_phone; p++) {
        if (noncomment_line(buf, sizeof(buf), fp) < 0)
            E_FATAL("Premature EOF reading phone %d\n", p);
        parse_tri_line(m, buf, p);
    }

    if (noncomment_line(buf, sizeof(buf), fp) >= 0)
        E_ERROR("Non-empty file beyond expected #phones (%d)\n",
                m->n_phone);

    /* Build CD senones to CI senones map */
    if (m->n_ciphone * m->n_emit_state != m->n_ci_sen)
        E_FATAL
            ("#CI-senones(%d) != #CI-phone(%d) x #emitting-states(%d)\n",
             m->n_ci_sen, m->n_ciphone, m->n_emit_state);
    m->cd2cisen = (int16 *) ckd_calloc(m->n_sen, sizeof(*m->cd2cisen)); /* freed in mdef_free */

    m->sen2cimap = (int16 *) ckd_calloc(m->n_sen, sizeof(*m->sen2cimap)); /* freed in mdef_free */

    for (s = 0; s < m->n_sen; s++)
        m->sen2cimap[s] = -1;
    for (s = 0; s < m->n_ci_sen; s++) { /* CI senones */
        m->cd2cisen[s] = s;
        m->sen2cimap[s] = s / m->n_emit_state;
    }
    for (p = n_ci; p < m->n_phone; p++) {       /* CD senones */
        for (s = 0; s < m->n_emit_state; s++) {
            cd = m->sseq[p][s];
            ci = m->sseq[m->phone[p].ci][s];
            m->cd2cisen[cd] = ci;
            m->sen2cimap[cd] = m->phone[p].ci;
        }
    }

    sseq_compress(m);
    fclose(fp);

    return m;
}
Exemple #2
0
/*
 * Initialize phones (ci and triphones) and state->senone mappings from .mdef file.
 */
mdef_t *
mdef_init(const char *mdeffile, int32 breport)
{
    FILE *fp;
    int32 n_ci, n_tri, n_map, n;
    char tag[1024], buf[1024];
    s3senid_t **senmap;
    /*    s3senid_t *tempsenmap; */

    s3pid_t p;
    int32 s, ci, cd;
    mdef_t *m;

    if (!mdeffile)
        E_FATAL("No mdef-file\n");

    if (breport)
        E_INFO("Reading model definition: %s\n", mdeffile);

    m = (mdef_t *) ckd_calloc(1, sizeof(mdef_t));       /* freed in mdef_free */

    if ((fp = fopen(mdeffile, "r")) == NULL)
        E_FATAL_SYSTEM("fopen(%s,r) failed\n", mdeffile);

    if (noncomment_line(buf, sizeof(buf), fp) < 0)
        E_FATAL("Empty file: %s\n", mdeffile);

    if (strncmp(buf, MODEL_DEF_VERSION, strlen(MODEL_DEF_VERSION)) != 0)
        E_FATAL("Version error: Expecing %s, but read %s\n",
                MODEL_DEF_VERSION, buf);

    /* Read #base phones, #triphones, #senone mappings defined in header */
    n_ci = -1;
    n_tri = -1;
    n_map = -1;
    m->n_ci_sen = -1;
    m->n_sen = -1;
    m->n_tmat = -1;
    do {
        if (noncomment_line(buf, sizeof(buf), fp) < 0)
            E_FATAL("Incomplete header\n");

        if ((sscanf(buf, "%d %s", &n, tag) != 2) || (n < 0))
            E_FATAL("Error in header: %s\n", buf);

        if (strcmp(tag, "n_base") == 0)
            n_ci = n;
        else if (strcmp(tag, "n_tri") == 0)
            n_tri = n;
        else if (strcmp(tag, "n_state_map") == 0)
            n_map = n;
        else if (strcmp(tag, "n_tied_ci_state") == 0)
            m->n_ci_sen = n;
        else if (strcmp(tag, "n_tied_state") == 0)
            m->n_sen = n;
        else if (strcmp(tag, "n_tied_tmat") == 0)
            m->n_tmat = n;
        else
            E_FATAL("Unknown header line: %s\n", buf);
    } while ((n_ci < 0) || (n_tri < 0) || (n_map < 0) ||
             (m->n_ci_sen < 0) || (m->n_sen < 0) || (m->n_tmat < 0));

    if ((n_ci == 0) || (m->n_ci_sen == 0) || (m->n_tmat == 0)
        || (m->n_ci_sen > m->n_sen))
        E_FATAL("%s: Error in header\n", mdeffile);

    /* Check typesize limits */
    if (n_ci >= MAX_S3CIPID)
        E_FATAL("%s: #CI phones (%d) exceeds limit (%d)\n", mdeffile, n_ci,
                MAX_S3CIPID);
    if (n_ci + n_tri >= MAX_S3PID)
        E_FATAL("%s: #Phones (%d) exceeds limit (%d)\n", mdeffile,
                n_ci + n_tri, MAX_S3PID);
    if (m->n_sen >= MAX_S3SENID)
        E_FATAL("%s: #senones (%d) exceeds limit (%d)\n", mdeffile,
                m->n_sen, MAX_S3SENID);
    if (m->n_tmat >= MAX_S3TMATID)
        E_FATAL("%s: #tmats (%d) exceeds limit (%d)\n", mdeffile,
                m->n_tmat, MAX_S3TMATID);

    m->n_emit_state = (n_map / (n_ci + n_tri)) - 1;
    if ((m->n_emit_state + 1) * (n_ci + n_tri) != n_map)
        E_FATAL
            ("Header error: n_state_map not a multiple of n_ci*n_tri\n");

    /* Initialize ciphone info */
    m->n_ciphone = n_ci;
    m->ciphone_ht = hash_table_new(n_ci, HASH_CASE_YES);  /* With case-insensitive string names *//* freed in mdef_free */
    m->ciphone = (ciphone_t *) ckd_calloc(n_ci, sizeof(ciphone_t));     /* freed in mdef_free */

    /* Initialize phones info (ciphones + triphones) */
    m->n_phone = n_ci + n_tri;
    m->phone = (phone_t *) ckd_calloc(m->n_phone, sizeof(phone_t));     /* freed in mdef_free */

    /* Allocate space for state->senone map for each phone */
    senmap = (s3senid_t **) ckd_calloc_2d(m->n_phone, m->n_emit_state, sizeof(s3senid_t));      /* freed in mdef_free */
    m->sseq = senmap;           /* TEMPORARY; until it is compressed into just the unique ones */


    /**CODE DUPLICATION!*****************************************************************************************************/
    /* Flat decoder-specific */
    /* Allocate space for state->senone map for each phone */

    /* ARCHAN 20040820, this sacrifice readability and may cause pointer
       problems in future. However, this is a less evil than
       duplication of code.  This is trick point all the state mapping
       to the global mapping and avoid duplicated memory.  
     */

    /* S3 xwdpid_compress will compress the below list phone list. 
     */

    /* ARCHAN, this part should not be used when one of the recognizer is used. */
    m->st2senmap =
        (s3senid_t *) ckd_calloc(m->n_phone * m->n_emit_state,
                                 sizeof(s3senid_t));
    for (p = 0; p < m->n_phone; p++)
        m->phone[p].state = m->st2senmap + (p * m->n_emit_state);
    /******************************************************************************************************/


    /* Allocate initial space for <ci,lc,rc,wpos> -> pid mapping */
    m->wpos_ci_lclist = (ph_lc_t ***) ckd_calloc_2d(N_WORD_POSN, m->n_ciphone, sizeof(ph_lc_t *));      /* freed in mdef_free */

    /*
     * Read base phones and triphones.  They'll simply be assigned a running sequence
     * number as their "phone-id".  If the phone-id < n_ci, it's a ciphone.
     */

    /* Read base phones */
    for (p = 0; p < n_ci; p++) {
        if (noncomment_line(buf, sizeof(buf), fp) < 0)
            E_FATAL("Premature EOF reading CIphone %d\n", p);
        parse_base_line(m, buf, p);
    }
    m->sil = mdef_ciphone_id(m, S3_SILENCE_CIPHONE);

    /* Read triphones, if any */
    for (; p < m->n_phone; p++) {
        if (noncomment_line(buf, sizeof(buf), fp) < 0)
            E_FATAL("Premature EOF reading phone %d\n", p);
        parse_tri_line(m, buf, p);
    }

    if (noncomment_line(buf, sizeof(buf), fp) >= 0)
        E_ERROR("Non-empty file beyond expected #phones (%d)\n",
                m->n_phone);

    /* Build CD senones to CI senones map */
    if (m->n_ciphone * m->n_emit_state != m->n_ci_sen)
        E_FATAL
            ("#CI-senones(%d) != #CI-phone(%d) x #emitting-states(%d)\n",
             m->n_ci_sen, m->n_ciphone, m->n_emit_state);
    m->cd2cisen = (s3senid_t *) ckd_calloc(m->n_sen, sizeof(s3senid_t));        /* freed in mdef_free */

    m->sen2cimap = (s3cipid_t *) ckd_calloc(m->n_sen, sizeof(s3cipid_t));       /* freed in mdef_free */

    for (s = 0; s < m->n_sen; s++)
        m->sen2cimap[s] = BAD_S3CIPID;
    for (s = 0; s < m->n_ci_sen; s++) { /* CI senones */
        m->cd2cisen[s] = (s3senid_t) s;
        m->sen2cimap[s] = s / m->n_emit_state;
    }
    for (p = n_ci; p < m->n_phone; p++) {       /* CD senones */
        for (s = 0; s < m->n_emit_state; s++) {
            cd = m->sseq[p][s];
            ci = m->sseq[(int) m->phone[p].ci][s];
            m->cd2cisen[cd] = (s3senid_t) ci;
            m->sen2cimap[cd] = m->phone[p].ci;
        }
    }

    sseq_compress(m);
    fclose(fp);

    return m;
}