IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
Enhanced AAC+, how to compare quality between original file and decode
happysaisai
post Sep 19 2012, 16:33
Post #1





Group: Members
Posts: 6
Joined: 19-September 12
Member No.: 103264



Hello,I'm doing with Enhanced aac+ codec, i met a problem ,is there someone knows which tool can be used to calculate the quality between original .wav file and the decoder output .wav file ?
I tried PEAQ, but the ODG value seems very low even though i did not add any error to the stream file.3gp.
Because the decoder output file is delayed, and the length between two files is different, so i shift the decoder output file and made the two files have same length, but still doesn't work, does someone know how to modify PEAQ or the two files?
Thanks.
Go to the top of the page
+Quote Post
greynol
post Sep 19 2012, 16:39
Post #2





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



http://www.hydrogenaudio.org/forums/index....showtopic=16295
http://wiki.hydrogenaudio.org/index.php?title=ABX

This post has been edited by greynol: Sep 19 2012, 16:40


--------------------
Concern trolls: not a myth.
Go to the top of the page
+Quote Post
C.R.Helmrich
post Sep 19 2012, 23:08
Post #3





Group: Developer
Posts: 686
Joined: 6-December 08
From: Erlangen Germany
Member No.: 64012



Regarding PEAQ:

  • Make sure both input (reference) and decoded (test) file are sampled at 48 kHz, since IIRC that's required by the PEAQ algorithm. Resample if necessary. You can encode from and to 44.1 kHz, though.
  • PEAQ mainly compares waveforms, so it doesn't give you any meaningful information on HE-AAC with its parametric stereo and bandwidth Extension (the ODGs are too low). See greynol's links.


Chris


--------------------
If I don't reply to your reply, it means I agree with you.
Go to the top of the page
+Quote Post
happysaisai
post Sep 20 2012, 09:02
Post #4





Group: Members
Posts: 6
Joined: 19-September 12
Member No.: 103264



QUOTE (C.R.Helmrich @ Sep 20 2012, 00:08) *
Regarding PEAQ:

  • Make sure both input (reference) and decoded (test) file are sampled at 48 kHz, since IIRC that's required by the PEAQ algorithm. Resample if necessary. You can encode from and to 44.1 kHz, though.
  • PEAQ mainly compares waveforms, so it doesn't give you any meaningful information on HE-AAC with its parametric stereo and bandwidth Extension (the ODGs are too low). See greynol's links.


Chris

Thanks.
both files are 48khz.
do you mean for the aac part, i can use PEAQ, but because Enhanced AAC+ includes the SBR+PS,so I can not use PEAQ ,right?
SO I can only use the subjective listening test , how about mushra test?
Go to the top of the page
+Quote Post
happysaisai
post Sep 20 2012, 09:04
Post #5





Group: Members
Posts: 6
Joined: 19-September 12
Member No.: 103264



QUOTE (greynol @ Sep 19 2012, 17:39) *

Thanks, but I can't open the 'Programs for Performing Blind Listening Tests' in this link http://wiki.hydrogenaudio.org/index.php?title=ABX
Go to the top of the page
+Quote Post
C.R.Helmrich
post Sep 20 2012, 10:12
Post #6





Group: Developer
Posts: 686
Joined: 6-December 08
From: Erlangen Germany
Member No.: 64012



QUOTE (happysaisai @ Sep 20 2012, 10:02) *
... how about mushra test?

Correct, for the AAC-LC part at high bitrates, you can use PEAQ, but you must average ODGs over many items to get reliable results. Yes, for SBR/PS, use a blind testing methodology like MUSHRA (if you have corresponding test software). ABC-HR is almost the same (but doesn't require a 3.5-kHz anchor condition) and works equally well.

Chris

This post has been edited by C.R.Helmrich: Sep 20 2012, 10:19


--------------------
If I don't reply to your reply, it means I agree with you.
Go to the top of the page
+Quote Post
happysaisai
post Sep 20 2012, 10:38
Post #7





Group: Members
Posts: 6
Joined: 19-September 12
Member No.: 103264



QUOTE (C.R.Helmrich @ Sep 20 2012, 11:12) *
QUOTE (happysaisai @ Sep 20 2012, 10:02) *
... how about mushra test?

Correct, for the AAC-LC part at high bitrates, you can use PEAQ, but you must average ODGs over many items to get reliable results. Yes, for SBR/PS, use a blind testing methodology like MUSHRA (if you have corresponding test software). ABC-HR is almost the same (but doesn't require a 3.5-kHz anchor condition) and works equally well.

Chris

Thanks again.smile.gif
1;could you please tell me how did you find the PEAQ is not good for SBR/PS?
2\do you know how to download the MUSHRA test software or ABC-HR?( i never did subjective listening test before) i followed the 'Programs for Performing Blind Listening Tests' in the link http://wiki.hydrogenaudio.org/index.php?title=ABX ,but i can not open it
3\ do you know how to use only AAC-LC part ,do not use SBR/PS?
My question seems too much. sorry to trouble you ,and thank you very much.
Go to the top of the page
+Quote Post
Garf
post Sep 20 2012, 14:18
Post #8


Server Admin


Group: Admin
Posts: 4884
Joined: 24-September 01
Member No.: 13



QUOTE (happysaisai @ Sep 20 2012, 11:38) *
1;could you please tell me how did you find the PEAQ is not good for SBR/PS?


By doing listening tests and seeing that the PEAQ output is completely uncorrelated with the result. As far as I can tell from your first post, you already found this yourself.

PEAQ records some crude information about the distortion and tries to match this to listening test results. If the distortion introduced is of a form PEAQ can't understand, or if the encoder is smarter and has a more advanced model of the human ear than PEAQ, it will give completely wrong results.

QUOTE
2\do you know how to download the MUSHRA test software or ABC-HR?


Try this one:
http://www.hydrogenaudio.org/forums/index....showtopic=32128

QUOTE
3\ do you know how to use only AAC-LC part ,do not use SBR/PS?


Perhaps that could be done by compiling a version of FAAD2, but disabling the HE-AAC decoding part of it.
Go to the top of the page
+Quote Post
greynol
post Sep 20 2012, 14:21
Post #9





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



If you are only comparing one lossy encoding against the original lossless source then you should use ABX instead of ABC/HR, though the later can be used like ABX but you have to conduct multiple trials in order to get any statistical significance.


--------------------
Concern trolls: not a myth.
Go to the top of the page
+Quote Post
happysaisai
post Sep 20 2012, 14:39
Post #10





Group: Members
Posts: 6
Joined: 19-September 12
Member No.: 103264



QUOTE (greynol @ Sep 20 2012, 15:21) *
If you are only comparing one lossy encoding against the original lossless source then you should use ABX instead of ABC/HR, though the later can be used like ABX but you have to conduct multiple trials in order to get any statistical significance.

thanks, but I can't download the ABX from the link ,do you have other way to download ?
Go to the top of the page
+Quote Post
happysaisai
post Sep 20 2012, 14:49
Post #11





Group: Members
Posts: 6
Joined: 19-September 12
Member No.: 103264



QUOTE (Garf @ Sep 20 2012, 15:18) *
QUOTE (happysaisai @ Sep 20 2012, 11:38) *
1;could you please tell me how did you find the PEAQ is not good for SBR/PS?


By doing listening tests and seeing that the PEAQ output is completely uncorrelated with the result. As far as I can tell from your first post, you already found this yourself.

PEAQ records some crude information about the distortion and tries to match this to listening test results. If the distortion introduced is of a form PEAQ can't understand, or if the encoder is smarter and has a more advanced model of the human ear than PEAQ, it will give completely wrong results.




thank you very much , i just want to ask where can I find the theory reason.

could you please tell me which listening tool do you use?
Go to the top of the page
+Quote Post
greynol
post Sep 20 2012, 15:07
Post #12





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



QUOTE (happysaisai @ Sep 20 2012, 06:39) *
QUOTE (greynol @ Sep 20 2012, 15:21) *
If you are only comparing one lossy encoding against the original lossless source then you should use ABX instead of ABC/HR, though the later can be used like ABX but you have to conduct multiple trials in order to get any statistical significance.

thanks, but I can't download the ABX from the link ,do you have other way to download ?

Have you tried searching the web?


--------------------
Concern trolls: not a myth.
Go to the top of the page
+Quote Post

Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 21st August 2014 - 15:19