TIMIT语音库

TIMIT语音库有着准确的音素标注，因此可以应用于语音分割性能评价，同时该数据库又含有几百个说话人语音，所以也是评价说话人识别常用的权威语音库，但该语音库的商业用途是要花钱买的。下面的资源来自与MIT教学实验使用，大概有430多M。

下载地址：http://web.mit.edu/course/6/6.863/share/nltk_lite/

不需要单个文件下载，可以使用下面的下载工具批量下载。

下载工具：http://www.onlinedown.net/soft/53010.htm

          The DARPA TIMIT Acoustic-Phonetic Continuous Speech Corpus
                                   (TIMIT)

                            Training and Test Data
                           NIST Speech Disc CD1-1.1




The TIMIT corpus of read speech has been designed to provide speech data for
the acquisition of acoustic-phonetic knowledge and for the development and
evaluation of automatic speech recognition systems.  TIMIT has resulted from
the joint efforts of several sites under sponsorship from the Defense Advanced
Research Projects Agency - Information Science and Technology Office
(DARPA-ISTO).  Text corpus design was a joint effort among the Massachusetts
Institute of Technology (MIT), Stanford Research Institute (SRI), and Texas
Instruments (TI).  The speech was recorded at TI, transcribed at MIT, and has
been maintained, verified, and prepared for CD-ROM production by the National
Institute of Standards and Technology (NIST).  This file contains a brief
description of the TIMIT Speech Corpus.  Additional information including the
referenced material and some relevant reprints of articles may be found in the
printed documentation which is also available from NTIS (NTIS# PB91-100354).




1. Corpus Speaker Distribution
-- ---------------------------

TIMIT contains a total of 6300 sentences, 10 sentences spoken by each of 630
speakers from 8 major dialect regions of the United States.  Table 1 shows the
number of speakers for the 8 dialect regions, broken down by sex.  The
percentages are given in parentheses.  A speaker's dialect region is the
geographical area of the U.S.  where they lived during their childhood years.
The geographical areas correspond with recognized dialect regions in U.S.
(Language Files, Ohio State University Linguistics Dept., 1982), with the
exception of the Western region (dr7) in which dialect boundaries are not
known with any confidence and dialect region 8 where the speakers moved around
a lot during their childhood.


   Table 1:  Dialect distribution of speakers

      Dialect
      Region(dr)    #Male    #Female    Total
      ----------  --------- ---------  ----------
         1         31 (63%)  18 (27%)   49 (8%) 
         2         71 (70%)  31 (30%)  102 (16%)
         3         79 (67%)  23 (23%)  102 (16%)
         4         69 (69%)  31 (31%)  100 (16%)
         5         62 (63%)  36 (37%)   98 (16%)
         6         30 (65%)  16 (35%)   46 (7%)
         7         74 (74%)  26 (26%)  100 (16%)
         8         22 (67%)  11 (33%)   33 (5%)
       ------     --------- ---------  ----------
         8        438 (70%) 192 (30%)  630 (100%)

The dialect regions are:
     dr1:  New England
     dr2:  Northern
     dr3:  North Midland
     dr4:  South Midland
     dr5:  Southern
     dr6:  New York City
     dr7:  Western
     dr8:  Army Brat (moved around)




2. Corpus Text Material
-- --------------------

The text material in the TIMIT prompts (found in the file "prompts.doc")
consists of 2 dialect "shibboleth" sentences designed at SRI, 450
phonetically-compact sentences designed at MIT, and 1890 phonetically-diverse
sentences selected at TI.  The dialect sentences (the SA sentences) were meant
to expose the dialectal variants of the speakers and were read by all 630
speakers.  The phonetically-compact sentences were designed to provide a good
coverage of pairs of phones, with extra occurrences of phonetic contexts
thought to be either difficult or of particular interest.  Each speaker read 5
of these sentences (the SX sentences) and each text was spoken by 7 different
speakers.  The phonetically-diverse sentences (the SI sentences) were selected
from existing text sources - the Brown Corpus (Kuchera and Francis, 1967) and
the Playwrights Dialog (Hultzen, et al., 1964) - so as to add diversity in
sentence types and phonetic contexts.  The selection criteria maximized the
variety of allophonic contexts found in the texts.  Each speaker read 3 of
these sentences, with each sentence being read only by a single speaker.
Table 2 summarizes the speech material in TIMIT.


    Table 2:  TIMIT speech material

  Sentence Type   #Sentences   #Speakers   Total   #Sentences/Speaker
  -------------   ----------   ---------   -----   ------------------
  Dialect (SA)          2         630       1260           2
  Compact (SX)        450           7       3150           5
  Diverse (SI)       1890           1       1890           3
  -------------   ----------   ---------   -----    ----------------
  Total              2342                   6300          10




3. Suggested Training/Test Subdivision
-- -----------------------------------

The speech material has been subdivided into portions for training and
testing.  The criteria for the subdivision is described in the file
"testset.doc".  THIS SUBDIVISION HAS NO RELATION TO THE DATA DISTRIBUTED ON
THE PROTOTYPE VERSION OF THE CDROM.


Core Test Set:

The test data has a core portion containing 24 speakers, 2 male and 1 female
from each dialect region.  The core test speakers are shown in Table 3.  Each
speaker read a different set of SX sentences.  Thus the core test material
contains 192 sentences, 5 SX and 3 SI for each speaker, each having a distinct
text prompt.


    Table 3:  The core test set of 24 speakers

     Dialect        Male      Female
     -------       ------     ------
        1        DAB0, WBT0    ELC0   
        2        TAS1, WEW0    PAS0   
        3        JMP0, LNT0    PKT0   
        4        LLL0, TLS0    JLM0   
        5        BPM0, KLT0    NLP0   
        6        CMJ0, JDH0    MGD0   
        7        GRT0, NJM0    DHC0
        8        JLN0, PAM0    MLD0   



Complete Test Set:

A more extensive test set was obtained by including the sentences from all
speakers that read any of the SX texts included in the core test set.  In
doing so, no sentence text appears in both the training and test sets.  This
complete test set contains a total of 168 speakers and 1344 utterances,
accounting for about 27% of the total speech material.  The resulting dialect
distribution of the 168 speaker test set is given in Table 4.  The complete
test material contains 624 distinct texts.


     Table 4:  Dialect distribution for complete test set

      Dialect    #Male   #Female   Total
      -------    -----   -------   -----
        1           7        4       11
        2          18        8       26
        3          23        3       26
        4          16       16       32
        5          17       11       28
        6           8        3       11
        7          15        8       23
        8           8        3       11
      -----      -----   -------   ------
      Total       112       56      168




4. CDROM TIMIT Directory and File Structure
-- ----------------------------------------

The speech and associated data is organized on the CD-ROM according to the
following hierarchy:

/<CORPUS>/<USAGE>/<DIALECT>/<SEX><SPEAKER_ID>/<SENTENCE_ID>.<FILE_TYPE>

     where,

     CORPUS :== timit
     USAGE :== train | test
     DIALECT :== dr1 | dr2 | dr3 | dr4 | dr5 | dr6 | dr7 | dr8
                 (see Table 1 for dialect code description)
     SEX :== m | f
     SPEAKER_ID :== <INITIALS><DIGIT>
         
          where,
          INITIALS :== speaker initials, 3 letters
          DIGIT :== number 0-9 to differentiate speakers with identical
                    initials
                             
     SENTENCE_ID :== <TEXT_TYPE><SENTENCE_NUMBER>
         
          where,
             
          TEXT_TYPE :== sa | si | sx
                        (see Section 2 for sentence text type description)
          SENTENCE_NUMBER :== 1 ... 2342
                   
     FILE_TYPE :== wav | txt | wrd | phn
                   (see Table 5 for file type description)

Examples:
     /timit/train/dr1/fcjf0/sa1.wav
                        
     (TIMIT corpus, training set, dialect region 1, female speaker,
      speaker-ID "cjf0", sentence text "sa1", speech waveform file)
     

      /timit/test/df5/mbpm0/sx407.phn
     
      (TIMIT corpus, test set, dialect region 5, male speaker, speaker-ID
       "bpm0", sentence text "sx407", phonetic transcription file)
     
                                                     
Online documentation and tables are located in the directory "timit/doc".
A brief description of each file in this directory can be found in Section 6.




5. File Types
-- ----------

The TIMIT corpus includes several files associated with each utterance.  In
addition to a speech waveform file (.wav), three associated transcription
files (.txt, .wrd, .phn) exist.  These associated files have the form:

        <BEGIN_SAMPLE> <END_SAMPLE> <TEXT><new-line>
        .
        .
        .
        <BEGIN_SAMPLE> <END_SAMPLE> <TEXT><new-line>

        where,       
       
                BEGIN_SAMPLE :== The beginning integer sample number for the
                                 segment (Note: The first BEGIN_SAMPLE of each
                                 file is always 0)
                                
                END_SAMPLE :== The ending integer sample number for the segment
                               (Note: Because of the transcription method used,
                               the last END_SAMPLE in each transcription file
                               may be less than the actual last sample in the
                               corresponding .wav file)

                TEXT :== <ORTHOGRAPHY> | <WORD_LABEL> | <PHONETIC_LABEL>
                        
                where,
               
                     ORTHOGRAPHY :== Complete orthographic text transcription
                     WORD_LABEL :== Single word from the orthography
                     PHONETIC_LABEL :== Single phonetic transcription code
                                        (See "phoncode.doc" for description
                                        of codes)



    Table 5:  Utterance-associated file types         

 File Type                     Description
 ---------  ------------------------------------------------------
    
     .wav - SPHERE-headered speech waveform file.  (See the "/sphere"
            directory for speech file manipulation utilities.)

     .txt - Associated orthographic transcription of the words the
            person said.  (Usually this is the same as the prompt, but
            in a few cases the orthography and prompt disagree.)

     .wrd - Time-aligned word transcription. The word boundaries
            were aligned with the phonetic segments using a dynamic
            string alignment program (see the printed documentation
            section "Notes on the Word Alignments" and the lexical
            pronunciations given in "timitdic.txt".)

     .phn - Time-aligned phonetic transcription.  (See the reprint
            of the article by Seneff and Zue (1988), in the printed
            documentation, and the section "Notes on Checking the
            Phonetic Transcriptions" for more details on the phonetic
            transcription protocols.)
            
                                       
Example transcriptions from the utterance in "/timit/test/dr5/fnlp0/sa1.wav"

Orthography (.txt):
        0 61748 She had your dark suit in greasy wash water all year.

Word label (.wrd):
        7470 11362 she
        11362 16000 had
        15420 17503 your
        17503 23360 dark
        23360 28360 suit
        28360 30960 in
        30960 36971 greasy
        36971 42290 wash
        43120 47480 water
        49021 52184 all
        52184 58840 year

Phonetic label (.phn):
(Note: beginning and ending silence regions are marked with h#)
        0 7470 h#
        7470 9840 sh
        9840 11362 iy
        11362 12908 hv
        12908 14760 ae
        14760 15420 dcl
        15420 16000 jh
        16000 17503 axr
        17503 18540 dcl
        18540 18950 d
        18950 21053 aa
        21053 22200 r
        22200 22740 kcl
        22740 23360 k
        23360 25315 s
        25315 27643 ux
        27643 28360 tcl
        28360 29272 q
        29272 29932 ih
        29932 30960 n
        30960 31870 gcl
        31870 32550 g
        32550 33253 r
        33253 34660 iy
        34660 35890 z
        35890 36971 iy
        36971 38391 w
        38391 40690 ao
        40690 42290 sh
        42290 43120 epi
        43120 43906 w
        43906 45480 ao
        45480 46040 dx
        46040 47480 axr
        47480 49021 q
        49021 51348 ao
        51348 52184 l
        52184 54147 y
        54147 56654 ih
        56654 58840 axr
        58840 61680 h#

           
           


6. Online Documentation
-- --------------------

Compact documentation is located in the "/timit/doc" directory.  Files in this
directory with a ".doc" extension contain freeform descriptive text and files
with a ".txt" extension contain tables of formatted text which can be searched
programmatically.  Lines in the ".txt" files beginning with a semicolon are
comments and should be ignored on searches.  The following is a brief
description of their contents:

    phoncode.doc - Table of phone symbols used in phonemic dictionary and
                   phonetic transcriptions
     prompts.txt - Table of sentence prompts and sentence-ID numbers
    spkrinfo.txt - Table of speaker attributes
    spkrsent.txt - Table of sentence-ID numbers for each speaker
     testset.doc - Description of suggested train/test subdivision
    timitdic.doc - Description of phonemic lexicion
    timitdic.txt - Phonemic dictionary of all orthographic words in prompts


A more extensive description of corpus design, collection, and transcription
can be found in the printed documentation.