
ISO/IEC 14496-4 MPEG-4 Audio Conformance Verification
SSNRCD - Segmental SNR / Cepstral Distortion Analysis Tool  
===============================================================


Software version history:

  Version nr.   Date             Authors        Description
--------------------------------------------------------------------------------------------
  1.0           26 Jan. 2000     RF             First released version.
  1.1           31 Jan. 2000     RF, TM         Cepstral distortion: only for segments
                                                with signal power in range [-50...-15] dB.
                                                Improved accuracy in calculation of 
                                                cepstral distortion.
  1.2           17 Apr. 2000     RF             MPEG-4 Version 2 Conformance criteria added.
  1.3           11 Oct. 2000     RF             Some compiler warnings removed on UNIX platforms.
  1.4           15 Nov. 2000     RF, RS         Delay compensation included.
                                                Object type based conf. criteria removed,
                                                -t <num> switch added instead. 
                                                K-bit conformance criteria included.
                                                General code clean-up.
  1.6           19 Jan. 2001     HP             Offset in conformance threshold due to
                                                rounding of PCM signals as introduced in
                                                version 1.4 removed again.
  1.8           20 March 2001    RF             -t 4 switch included
                                                General code clean-up.



RF   Ralf Funken (Philips)
TM   Takehiro Moriya (NTT)
RS   Ralph Sperschneider (FhG)
HP   Heiko Purnhagen (Hannover University)


This tool implements the MPEG-4 Conformance test for some of the 
MPEG-4 Audio audio objects. The test procedure is defined in 
ISO/IEC 14496-4 MPEG-4 Conformance. Calculated are the average 
segmental SNR, the RMS value of the difference signal, the 
maximum absolute difference and the cepstral distortion.


Usage:   ssnrcd  [options] <reference inputfile> <inputfile under test>

Valid options are:
                -d <num>     Time-offset in samples between reference inputfile and
                             inputfile under test. Positive value means inputfile under test
                             is delayed over <num> samples and the first <num>
                             samples of that input file under test are skipped
                             Negative values are permitted.
                             Default: 0 samples.
                -k <num>     Resolution for calculation of RMS and max. abs.
                             difference sample, expressed in bits.
                             Valid range: [12...24]  Default: 16 bits.
                -l <num>     Analysis segment length in samples.
                             Default: 320 samples.
                -t <num>     Conformance criteria to be applied.
                               0 : No criterion (default)
                               1 : RMS + Max. absolute difference sample
                                   using K bit resolution (see -k option)
                               2 : Segmental SNR + avg. Cepstral Distortion
                               4 : Number of bits accuracy reached for RMS criterion
                             These values may be added. For example -t 7 enables
                             all three criteria.
                -v           Verbose progress report.

If the input files have different length, only the number of samples
in the shortest file will be used. The additional samples at the end of
the longest file will be ignored.

Segmental SNR and cepstral distortion are calculated only for
segments with reference signal power in the range [-50...-15] dB.
Any partial segments at end of input signal are padded with
zeroes when calculating segmental SNR and Cepstral Distortion.
The length of the segments can be set by using the -l option.

Both input files must have the same number of audio channels and the same
sampling rate. Audio files in wav, aiff, aifc and au format are accepted.
The maximum number of audio channels is 64.

