readme.txt

实现fs1016w的CELP的低速率语音编解码功能的基于vc开发环境的原代码。

README AND QUICK REFERENCE
US FEDERAL STANDARD - 1016 CELP v3.2 FOR MATLAB (FS-1016)

DEVELOPED BY TED PAINTER AND ANDREAS SPANIAS
COPYRIGHT (C) 1994-99 ANDREAS SPANIAS AND TED PAINTER

THE AUTHORS GRATEFULLY ACKNOWLEDGE THE MANY CONTRIBUTIONS OF THE SPEECH
CODING GROUP AT NSA.  IN ADDITION TO RECOGNIZING THEIR OUTSTANDING TECHNICAL  
ACHIEVEMENTS, WE PARTICULARLY THANK THEM FOR MAKING AVAILABLE 'C' 
SOURCE CODE FOR THE FEDERAL STANDARDS.

2/8/99

1.0  SYSTEM REQUIREMENTS

  * 486DX4-100 or better 
  * 16+ MB recommended.
  * 512 kb free hard disk space minimum
  * Microsoft-compatible mouse
  * Microsoft Windows v3.1
  * MATLAB for Windows v4.0 or better (also runs under the Student Edition)
  * Software will also run under MATLAB in UNIX and other workstation environments.

2.0  INSTALLATION

  a)  PLEASE READ THE COPYRIGHT NOTICE PRIOR TO INSTALLATION.  IT CAN BE FOUND
      IN THE FILE CPYRIGHT.TXT.
  c)  Copy fs1016w.exe to a directory in which you wish to keep the software.
  d)  Type "fs1016w".  Press Return.  This will unpack the fs1016 archive.
  e)  Delete the archive file fs1016w.exe.
  f)  Follow the instructions under 3.0 to run the software.
  g)  Refer to section 4.x for technical notes and references.

3.0  RUNNING FS-1016 FOR MATLAB 

FS-1016 for MATLAB uses Windows-standard .WAV files for speech input and 
speech output.  Input files should be 16-bit .WAV files, sampled at 8 kHz.  Most 
sound card software will support this file format.  Similarly, output files generated 
by the algorithm are Windows-standard 16-bit .WAV files, with an 8 kHz sample 
rate.  Follow these steps to run the software:

  a)  Run MATLAB for Windows.
  b)  Change to the FS-1016 directory, e.g., "cd \fs1016".  Press Return.
  c)  Type "fs1016".  Press Return.
  d)  Read and acknowledge the copyright notice.
  d)  Answer the input and output file selection dialogs with appropriate input and 
      output files.
  e)  In addition to post-filtered output, the simulations generate both highpass
      and non-postfiltered outputs as well.  The .cha file contains an ASCII 
      hex-encoded representation of each set of frame parameters.

IMPORTANT NOTE:

PLEASE READ THE COPYRIGHT NOTICE IN THE ALGORITHM DIRECTORY BEFORE
INSTALLING OR USING THE SPEECH CODING SOFTWARE.
MATLAB is a trademark of The MathWorks.

4.0  TECHNICAL NOTES (MOSTLY EXCERPTED FROM FS-1016 C-SOURCE, COPYRIGHT US GOVERNMENT)

 4.1 REFERENCES

     Tutorials:
        Fallside, Frank and William Woods, Computer Speech Processing,
        Prentice Hall International, 1985, Chapter 4 (by Bishnu Atal).
 
        Spanias, A., "Speech Coding:  A Tutorial," Proc. IEEE, 1994. 

        Spanias, A., and Painter, T., "A software tool for introducing 
        speech coding fundamentals in a DSP course," IEEE Trans. Educ. v.39, 
        no.2, p143-52, May 1996.

      FS-1016:
        National Communication System Technical Information Bulletin
        Federal Standard 1016 (to be published 1992).
 
        Campbell, Joseph P. Jr., Thomas E. Tremain and Vanoy C. Welch,
        "The Federal Standard 1016 4800 bps CELP Voice Coder," Digital
        Signal Processing, Academic Press, Vol1, No. 3, p. 145-155.
 
        Kemp, David, P., Retha A. Sueda and Thomas E. Tremain, "An
        Evaluation of 4800 bps Voice Coders," Proceedings of the IEEE
        International Conference on Acoustics, Speech and Signal Processing
        (ICASSP), 1989, p. 200-203.
 
        Fenichel, R., "Federal Standard 1016," National Communications
        System, Office of Technology and Standards, Washington, DC 20305-2010,
        14 February 1991.
 
        Campbell, Joseph P. Jr., Thomas E. Tremain and Vanoy C. Welch,
        "The DoD 4.8 kbps Standard (Proposed Federal Standard 1016),"
        "Advances in Speech Coding", Kluwer Academic Publishers, 1991,
        Chapter 12, p. 121-133.


 4.2 4800 bps CELP Characteristics
 
                   Spectrum       Pitch            Code Book
                   -------------  ---------------  -----------------
        Update     30 ms          30/4 = 7.5 ms    30/4 = 7.5 ms
                   ll=240         lp=60            l=60
 
        Order      10             256 (max) x 60   512 (max) x 60
                                  1 gain           1 gain
 
        Analysis   Open loop      Closed loop      Closed loop
                   Correlation    Modified MSPE    MSPE VQ
                   30 ms Hamming  VQ, weight=0.8   weight=0.8
                   no preemphasis range 20 to 147  shift by 2
                   15 Hz BW exp   (w/ fractions)   77% sparsity
 
        Bits per   34 indep LSP   index:  8+6+8+6  index:     9*4
        Frame     [3444433333]   gain(-1,2): 5*4  gain(+/-): 5*4
 
        Bit Rate   1133.3 bps     1600 bps       1866.67 bps
 
        NOTE:  The remaining 200 bps are used as follows:  1 bit per frame
        for synchronization, 4 bits per frame for forward error correction
        and 1 bit per frame to provide future expansion(s) of the coder.
 

 4.3. UNPERMUTED BIT ASSIGNMENT
 
        lsp  1  1-3             lsp  6  20-22
        lsp  2  4-7             lsp  7  23-25
        lsp  3  8-11            lsp  8  26-28
        lsp  4  12-15           lsp  9  29-31
        lsp  5  16-19           lsp 10  32-34
 
        Subframe:         1       2       3        4
        ----------      -----   -----   -----    -------
        pitch delay     35-42   .....   87-94    .......
        delta delay     .....   62-67   .......  114-119
        pitch gain      43-47   68-72   95-99    120-124
        cbindex         48-56   73-81   100-108  125-133
        cbgain          57-61   82-86   109-113  134-138
 
        future bit      139
        error control   140-143
        sync            144
 
                FORWARD ERROR CONTROL, HAMMING (15,11):
        bits protected          allocation      rate    parameter
                                0               1       LSP
        40,41,42,92,93,94       1st-3rd MSBs    2       Pitch delay
                                0               2       Delta pitch delay
        47,72,99,124            1st MSB         4       Pitch gain
                                0               4       CB index
                                0               4       CB gain
        139                     1               1       Future bit
 
                SYNCHRONIZATION BIT
        The sync bit (144) begins with 0 in the first frame, then alternates 
        between 1 and 0 on successive frames.