I'm new here, so I may not know how things go around here but anyway.
I'm doing my university project on objective quality tests of compressed audio, using different tools mentioned in title. But I have few minor issues.
With the exception of PEAQ, I'm getting some strange results (quite low ODG, even with good, high quality codecs). In case of PEMO-Q I had none, because the algorithm (or its implementation) couldn't process them.
First idea was to align the files, since some compressed files were shifted in time. I've done that in two different ways (Audition - Auto speech alignment; and Audacity - manually aligned). From this point I'm testing only on EAQUAL (PEAQ and PEMO-Q I can only access at school) - and I didn't get much better results. From Audition it was even worse, Audacity somewhat better, but still far from subjective and PEAQ results.
So now I'm here, wondering what else I could do, since the alignment shouldn't be a problem (most likely). Do you have any ideas, what could I do with files before handing them to algorithm? Any ideas would be appreciated. I can hand out some of the testing files and/or the results if necessary. Thanks in advance.
When it comes to alignment, I highly suggest Pluraleyes. I use it in a video environment but it will sync and correct for any drift (before you freak out, it's really minor and usually a video-only issue). You may need to use an NLE (such as Premiere) as a middleman to export them so they have identical duration and alignment.
They have a free trial and an academic price.