R.wav - reference audio PCM
T.wav - test audio PCM
Any criteria or formula to show the quality of T.wav refer to R.wav ?
Full Version: Criteria to decide audio quality
QUOTE
It is not proof of quality, but it is proof of equal or disparate quality.
ABX seems not to tell how good the T.wav is.
Especially while there are two test files, e.g. T1.wav and T2.wav.
What you are looking for is an ABC/HR testing tool like ff123's.
It depends on what you mean with "quality".One method could be a plain file compare:
diff (or fc.exe) T.wav R.wav
If the files are equal then T.wav has 100% quality, otherwise it is crap.
More sophisticated methods could be:
-subtract T.wav from R.wav (sample-by-sample) and calculate the power of the difference.wav
-perform listening (blind) tests to find out if there are any audible differences (see other posts for links)
AFAIK, some high quality audio codecs always use bit-lost/bit-precision to show how good quality they are.
I wonder it's the only method for measuring audio quality.
Is there anything else? If yes, what is the methodology?
I wonder it's the only method for measuring audio quality.
Is there anything else? If yes, what is the methodology?
That depends on what you mean by quality. Better quality could mean less encoded errors in a file vs another or it could be subjective quality where you believe something sounds better than something else but in reality it could have more errors than the file it's being tested against.
QUOTE(ChangFest @ Jul 23 2004, 02:48 PM)
... or it could be subjective quality where you believe something sounds better than something else but in reality it could have more errors than the file it's being tested against.
So subjective and objective tests are both required to judge the quality of audio files ?
Or a objective test dominates the quality?
Please describe more precisely what it is you wish to do.
If you want to compare the quality of an mp3 encode to the original then there is no exact mathematical method. EAQUAL tries to do this but is far from perfect.
EAQUAL is objective in the sense that with the same input both you and I will obtain the exact same results whereas our opinions might differ strongly once we actually listen to the sample.
In terms of audio, quality can be thought of as "comfort". You could give an exact objective description of a chair but one of us might find the chair more comfortable than the other. The problem is that objective and subjective don't neccessarily correlate and in our case the subjective impression is the most important one.
If you want to compare the quality of an mp3 encode to the original then there is no exact mathematical method. EAQUAL tries to do this but is far from perfect.
EAQUAL is objective in the sense that with the same input both you and I will obtain the exact same results whereas our opinions might differ strongly once we actually listen to the sample.
In terms of audio, quality can be thought of as "comfort". You could give an exact objective description of a chair but one of us might find the chair more comfortable than the other. The problem is that objective and subjective don't neccessarily correlate and in our case the subjective impression is the most important one.
AFAIK, some audio decoders have the following requirements to their decoders' implementation.
1. absolute value of the samples, ref(n) and test(n), couldn't excess N. (bit precision)
2. maximum value of the noise(difference) to signal ratio couldn't greater than E (db).
I doubt evev one implements a decoder with bit loss M, for example, equals N+3.
The listeners have the ability to tell which is better than another ?
So i would like to know if there are another methods out of there?
1. absolute value of the samples, ref(n) and test(n), couldn't excess N. (bit precision)
2. maximum value of the noise(difference) to signal ratio couldn't greater than E (db).
I doubt evev one implements a decoder with bit loss M, for example, equals N+3.
The listeners have the ability to tell which is better than another ?
So i would like to know if there are another methods out of there?
What do you want to compare?
Bit precision is used as a measure for fixed-point implementations of e.g. decoders - compared to a reference implementation.
Or do you need a perceptual measure?
Then there's e.g. http://www.opticom.de who seem to have quality analyzers based on perceptual models.
Bit precision is used as a measure for fixed-point implementations of e.g. decoders - compared to a reference implementation.
Or do you need a perceptual measure?
Then there's e.g. http://www.opticom.de who seem to have quality analyzers based on perceptual models.
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.