To all those who did the tests and sent me the results, thank you very much, to those who downloaded the files with the intent to do the tests thanks very much to you too.
My area of research is sub-frequency bandwidth extension for live performance, which means creating sub-harmonics from a signal and mixing them back in with the original. This effectively extends the bandwidth of a signal by say an octave. I am at a stage where I need to know if what I am developing is having a significant effect or not... in other words if people can hear a difference between one version and another.
The main reason for putting these tests up here is to have a dry run with people that will probably listen with a more critical ear than Joe Public. The tests as they stand are in part fundamentally flawed, since the tones being compared are not always at the same level within the same sub-test (it's a gain symptom from the adaptive filtering I've designed) and people are more likely to say that a louder tone is better quality than a quieter one. But, nevertheless, the main part of the test was to establish that the adaptive filtering for each algorithm is having a significant effect on removing all the unwanted parts (frequency components) of the signal and getting close to a pure reference tone of the same pitch. The results showed overwhelmingly this to be the case, since the filtered outputs were also those with the slightly lower levels and were selected as superior to the unfiltered version, with good quality ratings with respect to the reference. This means that the filtering is doing its job.
The other thing I wanted to get from these tests is to establish if anyone could hear significant distortions ..... most people didn't (some did so well done) mention the slow attack times on the waveform due to the very narrow band filtering... which is interesting. Some mentioned more colouration in the "better" tone compared to the reference, which is good information for me.
I also learnt a lot about how to put these tests together, and in some cases, how not to do it!! GuitarGeorge was the first person to do the test (despite having an ear infection at the time) and provided some very useful tips and information. Many thanks George.
I will be putting a much easier test (in terms of doing it) up on YouTube in the near future, which will address all the points of this one so the test is as flawless as possible in the absence of controlled conditions (including constant level). The YouTube environment doesn't seem to have a significant effect on the audio uploaded, I tried a test version today. Hopefully, with that in place I can get a subject quota of about 100.
i'll put a link up for the YouTube video, when it appears. If you could also do that test, that would be really cool, it's about a 5 min video, so with a few pauses t will take 10 minutes max. The tests run sequentially and you can use the pause button to stop it at any time. I'll post the answer template in the video description so a quick copy and paste to a text file will sort that. The questions themselves are part of the video.
Many thanks ladies and gents..
Mike
