/ Data / Fundamentals / Production / Tutorials / Software / Home
2.4.1 Auxiliary Resources: TIDigits
Section 2.4.1: The TIDigits Database
The TI Digits database was one of the first publicly available databases used in speech research. It was collected at Texas Instruments in the early 1980's:
  • Authors: R. Gary Leonard and George Doddington
  • Data type: speech
  • Sample rate: 20000 Hz
  • Sampling format: 1 channel 16-bit linear compressed
  • Data source: microphone
  • Application: speech recognition
  • Language: English
Section 2.4.1: The TIDigits Database
This three-disc set contains speech which was originally designed and collected at Texas Instruments, Inc. (TI) for the purpose of designing and evaluating algorithms for speaker-independent recognition of connected digit sequences. There are 326 speakers (111 men, 114 women, 50 boys and 51 girls) each pronouncing 77 digit sequences. Each speaker group is partitioned into test and training subsets.

The database is available from LDC. Click here for more details. A subset of this database is included in our software release and used for validation purposes. This is a good database for testing basic features of a software implementation because it provides rapid turnaround time due to its small size and small vocabulary, and yet represents a non-trivial recognition task.

Click here to download the subset of the TIDigits database that we will use for this tutorial. This subset consists of 941 files used for training a system and 336 files used for evaluating a system. We will make use of this dataset throughout this tutorial. Note that this data is in compressed gzip format, and needs to be unpacked using the tar command described in Section 1.4.1.

Go to the directory:
    $ISIP_TUTORIAL/sections/s02/s02_04_p01/
After unpacking this data, you can verify the format of the data using this command:

    od -t d2 tidigits_8k_raw/test/man/fa/fa_3b.raw | head -3
which should produce this output:

    0000000 -1 -2  0  0  0  0  0  0
    0000020  0 -1 -1  0  0  0 -1  0
    0000040 -2  0 -1  0  0 -1  0  0
If your output does not look like this, the data in RAW files may need to be byte-swapped. This can be done with SoX using the following command:

     sox -t .sw tidigits_8k_raw/test/man/fa/fa_3b.raw -x -t .sw tidigits_8k_raw/test/man/fa/fa_3b_swap.raw

For a description of other databases represented in our environment, see Support of Industry-Standard Databases as well as Section 7.
   
Table of Contents   Section Contents   Previous Page Up Next Page
      Glossary / Help / Support / Site Map / Contact Us / ISIP Home