Blind Audio

2002-01-01 19:56:43

Hey Sphoid,

How's the tool looking?  Can you put up a screen shot?  Is there something in rudimentary form that is usable?  Here's a picture of what might be useful.  I don't know what you've currently got planned:



For each sample there are two slide bars.  The user rates both the sample and the original each time.  A and B are the sample or the original in random order.  The green buttons on top are combination play/stop buttons.  Of course, you'd probably have 8 pairs of buttons to handle the max number of samples, but ideally the window would dynamically size itself to the test.  Every time a listener loads a test configuration file, the samples would be loaded and randomized.  The ratings are stored in another file.  The answer key is saved along with the ratings, preferably in encrypted format.  It doesn't have to be some super-duper encryption; it could be ROT13 for all I care, but it should be good enough to discourage the listener from peeking at the answers and influencing him to either retry the test to get better results or to not send in his results because he thinks they're too poor.

I really don't want to organize another listening test unless there is some good way to filter out noisy listeners, and a tool like this would probably be a great help for that.

ff123

Edit:  Oops -- I forgot an important part of the tool.  It needs text entry boxes for comments!

Edit 2:  Probably all the sliders should start from 5.0, if either A or B is slid to a position less than 5.0, the other slider must automatically be at 5.0 (Either A or B is the original!)

Blind Audio

Reply #1 – 2002-01-01 23:06:12

Sorry, no screen shots yet since i pretty much started from scratch. i should have something soon. Ive got a usable version right now but its missing alot of features right now so im not going to release it quite yet. Your diagram will be very useful, thank you. Ill probably post some updates later on it.

Blind Audio

Reply #2 – 2002-01-08 01:16:04

Another way to make sure that the results aren't noisy is to invite a panel of "trusted" listeners to evaluate a lot of different samples.  For example, what would it be worth to see the ratings of a panel consisting of:

xiphmont, dibrom, johnv, filburt, hans heijden, david robinson, ivan, etc.

on say, a dozen different samples?  Probably worth quite a bit, is my guess.

ff123

Blind Audio

Reply #3 – 2002-01-20 03:31:55

Here's what Sphoid and I are looking at right now:



It's similar to the layout ff123 suggested except that instead of a seperate group of panels per sample on the same page, you just select between the sample sets via the top slider.  The "results so far" panel will show some sort of relational graph with the current ratings the user as chosen for each sample.

This is a pretty early layout, and things will probably get more advanced as they progress, but it should provide the basics as it is.  I think we'll have a beta version out in a few days to a weeks time.

Blind Audio

Reply #4 – 2002-01-24 07:54:47

Updated the picture to include the graph system... things are still coming along.