[Mp4-tech] mpeg 4 PQevalAudio evaluation tool

akhalife uci.edu akhalife uci.edu
Wed Mar 28 01:17:14 EDT 2007


Dear folks,
I am using the AFsTool, to evaluate the performance of the wav files,my
first question, does any body recommend other performance measurements
tools?
Also I am facing the follwoing problem(gain and time alignements), so if
anybody knows about it, please respond.
Thanks
Ala
here is the description
Currently I am using the MPEG 4 AFS tools  for evaluating the speech
quality, more  specifically (PQevalAudio).
>> But I am confused from the results (I mean the Objective Difference
>> grade).  after reading its html help file, I have noticed that I have
to align the  reference and the modified wav files, so you mentioned
that that can be  done using the CompAudio tool.
>>
>> so I did that ,and below is what I did:
>> 1)first ./CompAudio /testdir/s2.wav /testdir/s2-modified.wav  WAVE file:
>> /root/mpeg4/audio/natural/test-dir/test1/s2.wav
>>   Number of samples : 1024 (0.02133 s)  2024-03-27 02:58:31 UTC
>>   Sampling frequency: 48000 Hz
>>   Number of channels: 1 (16-bit integer)  WAVE file:
>> /root/mpeg4/audio/natural/test-dir/test1/s2-modified.wav
>>   Number of samples : 1024 (0.02133 s)  2024-03-27 03:20:41 UTC
>>   Sampling frequency: 48000 Hz
>>   Number of channels: 1 (16-bit integer)
>>
>> File A:
>>    Number of Samples: 1024
>>    Std Dev = 55.037 (0.168%),  Mean = 0.4043 (0.001234%)
>>    Maximum = 157 (0.4791%),  Minimum = -195 (-0.5951%)  File B:
>>    Number of Samples: 1024
>>    Std Dev = 0.054074 (0.000165%),  Mean = -0.0029297 (-8.941e-06%)
>>    Maximum = 0 (0%),  Minimum = -1 (-0.003052%)
>> SNR      = 0.00043307 dB
>> SNR      = 0.011383 dB  (Gain for File B = 52)
>> Seg. SNR = 0.001003 dB  (768 sample segments)  Max Diff = 195
>> (0.5951%),  No. Diff = 1011 (7 runs)
>>
>> 2)Now how to align the gain and the time differece?  here is what I did
>> ?
>> ./CopyAudio -g 52 /testdir/s2.wav /testdir/s2-modified.wav
>>
>> 3)for the time alignmenet? I do not know, is these aligned, and if not,
>> what to do?
>> 4)then I used the ./PEQ tool, but I got low score (means that the two
>> files are similar), knowing that they are not similar, since I
>> introduced errors to the second file, while before coping the gain, the
>> score was around -3.7 (very annoying). so can you exaplain more what do
>> you mean by gian and time difference alignment using the CopyAudio tool.
>>
>> Thanks alot
>> Ala
>>
>>



More information about the Mp4-tech mailing list