My Psychology Project

[avaserfi]

So I have a research methodology course who's basis is psychological research this year. We are finishing up our whole class assignment and are going to be given free reign to come up with a research question of our own. Naturally I thought of this hobby and all the various tests that can be done. I figured a college campus would be a great population to work with as most people I would include likely spend most of their time listening to compressed and completely dynamic-less music.

I am trying to come up with a good research question that I can apply to a smallish sample size. I will likely try to have n=50 or so. More if there is time.

So far I have a couple I would like to test I just need to decide:

1) I could go the easy route and have people listen at SPL X, SPL X+5, SPL X+10 to see which is more liked.
2) I could use a very dynamic song and compare to the same song but compressed as done in the loosing battle the loudness wars.

As you can see I haven't put my full thought into this yet, but am in a preliminary planning stage. Any other suggestions/opinions would be great as I have some time before my proposal is due. Currently I am trying to focus that can be done on portable media as I carry my laptop with me to class and this would allow me to get a larger sample size. The only "problem" is I would need to get a nice pair of headphones for the job .

 

[Adam]

You should try to find a way to fit fancy power cords or ERS paper into it.

 

[Joe Schmoe]

You should try to find a way to fit fancy power cords or ERS paper into it.

I think he is looking for a test that will actually produce results.
I like the compressed vs uncompressed concept, but where do you get both in the same format? (Comparison using two formats fails to isolate compression as the variable.)

 

[avaserfi]

I think he is looking for a test that will actually produce results.
I like the compressed vs uncompressed concept, but where do you get both in the same format? (Comparison using two formats fails to isolate compression as the variable.)

I have audio editing software so I can take a quality recording and compress it myself at different levels all in a losseless format. This would allow me to compare no compression, compression X, compression greater than X etc...Thus allowing for a possible threshold to emerge as well.

I would likely have to do some research on proper compression methods to make sure I do so appropriately.

One idea that was mentioned to me that I like is a comparison of dynamic versus dynamically compressed samples at the same perceptual SPLs, but am unsure of a proper way to achieve the same perceptual SPLs from both samples. Any thoughts/ideas?

 

[zhimbo]

SPL matching is a tricky issue with with varying the compression levels, isn't it? If you match on average SPL, the peaks are different, if you match the peaks the average is different.

If you're testing "preferences" you could bypass that issue by letting people adjust volume for each sample from "0" to their preferred level for the sample, to find which sample they prefer listening to when they have full control of the volume level. Pretty good "real world" conditions, actually. I'd use headphones with good isolation so they aren't fighting background noise.

So, does changing the bitrate of mp3s change the dynamic range? Probably, now that I think about it, but if not, this would be an easier thing to test using constant SPL.

 

[avaserfi]

SPL matching is a tricky issue with with varying the compression levels, isn't it? If you match on average SPL, the peaks are different, if you match the peaks the average is different.

If you're testing "preferences" you could bypass that issue by letting people adjust volume for each sample from "0" to their preferred level for the sample, to find which sample they prefer listening to when they have full control of the volume level. Pretty good "real world" conditions, actually. I'd use headphones with good isolation so they aren't fighting background noise.

So, does changing the bitrate of mp3s change the dynamic range? Probably, now that I think about it, but if not, this would be an easier thing to test using constant SPL.

That is the exact problem SPL matching would be extremely tricky hence why I might not be able to do this experiement. The problem with allowing every person to adjust the volume as needed is that it will introduce another variable that could bias the results. Also, it is well known that higher SPLs are perceived as better so if I do not have some sort of method to match SPLs, at least the average output, to attempt to control for this problem.

Changing the bitrate would effect the dynamic range, but I was planning a more controlled method of taking a lossless format then controlling the amount of dynamic compression by certain factors using Sound Forge.

Lastly, if I do this I will be purchasing a relatively good pair of headphones with as linear a frequency response as possible given my budget to control as much of the environment as possible as well as not have the speaker type bias results.

 

[gmichael]

You could get two pair of headsets. One Bose, one Kosh. Blindfold them and put them on one at a time playing the same music. See how many pick the Bose. Then, take the blindfolds off and do it again. I bet that more people will pick Bose the second time around. Go from there.

 

[avaserfi]

You could get two pair of headsets. One Bose, one Kosh. Blindfold them and put them on one at a time playing the same music. See how many pick the Bose. Then, take the blindfolds off and do it again. I bet that more people will pick Bose the second time around. Go from there.

That would be a simple experiment but there is no way in hell I am putting my hard earned money towards anything with the bose name on it. Also, it is pretty well known that preference lies with recognized brands. I am trying to do something more on the audio side and less on the brand recognition side.

 

[WmAx]

I can prepare the perceptual samples for you. Just provide me with lossless sample files for me to work from.

-Chris

 

[Audioholics]

Compression is just good clean fun. I'd like to know how 50 people reacted. I'd also make them describe the sound of each track using a few words.

 

[avaserfi]

Compression is just good clean fun. I'd like to know how 50 people reacted. I'd also make them describe the sound of each track using a few words.

That is a very good idea. I was planning on having a questionnaire to find out why the subjects enjoyed A more than B and visa versa that kind of description would be pretty interesting.

I just need to find a reasonable way to implement this idea and attempt to eliminate this SPL bias. The more I think about it the more willing I am to set the money aside to buy a good set of cans. I need a pair anyways .

 

[avaserfi]

So I found a program that claims to match perceived loudness between different music files. The process seems to make sense to me, but I was hoping that someone with more knowledge, as I admittedly have little in this area, could look over their process and see if it is valid.

Home Page: http://replaygain.hydrogenaudio.org/
How it works: http://replaygain.hydrogenaudio.org/contents.html

 

[avaserfi]

I have had Adobe Audition brought to my attention and its ability to equalize percieved SPL between dynamic and dynamically compressed sources. So it seems as if the research proposal to study the preference of listeners between these variables will be a possibility.

Now to find a quality closed headphone with a linear response that is around $100-150. Right now the front runner is Sony MDR-7506.

 

[bwaslo]

Why not let the listeners set the volume level for each themselves? That's what they do when they listen to compressed/uncompressed sound on their own, it would be a more fair test -- which do they like better when they can adjust either to any volume level they want?

 

[avaserfi]

Why not let the listeners set the volume level for each themselves? That's what they do when they listen to compressed/uncompressed sound on their own, it would be a more fair test -- which do they like better when they can adjust either to any volume level they want?

There are a few fundamental problems with that suggestion:

1) Introducing variability as different users listen at different levels
2) It has been proven time and time again that louder is perceived as better meaning if the two levels are not matched perceptively they dynamic source will be considered superior because it has been proven that dynamic sources are perceived as louder by individuals
3) I would not want subjects changing volume mid song as that would defeat the purpose of the study especially if both the dynamic version and compressed version are heard at different volumes
4) Remember we want to compare the control (uncompressed) to various amounts of dynamic compression. If the control and compressed songs are not heard in the same context how can a comparison be made?

If you would like citations please PM me as I have a majority of these articles on my computer, but please do look into the subject of perception if you are interested as these suggestions you make may seem logical, but in reality add great bias to the experiment which would make it completely ungeneralizable and thus useless.

 

[OttoMatic]

Now to find a quality closed headphone with a linear response that is around $100-150. Right now the front runner is Sony MDR-7506.

If you'd consider an open headphone, I would highly recommend the Grado offerings. Extremely nice stuff, and well-received by many.

I'm not familiar with the SONY cans, but I'd never buy another anything that says SONY on it. Just my opinion there.

I think your idea of comparing Bose to others is an interesting one. I understand that you wouldn't want to pay for them. I think you could buy them and then take them up on whatever money-back guarantee they have.

 

[avaserfi]

If you'd consider an open headphone, I would highly recommend the Grado offerings. Extremely nice stuff, and well-received by many.

I'm not familiar with the SONY cans, but I'd never buy another anything that says SONY on it. Just my opinion there.

I think your idea of comparing Bose to others is an interesting one. I understand that you wouldn't want to pay for them. I think you could buy them and then take them up on whatever money-back guarantee they have.

That actually wasn't my idea I believe it was gmicheal's these are my thoughts on the subject:

That would be a simple experiment but there is no way in hell I am putting my hard earned money towards anything with the bose name on it. Also, it is pretty well known that preference lies with recognized brands. I am trying to do something more on the audio side and less on the brand recognition side.

I am going to be getting over ear phones so they will be useful later on personally and also because I might not be able to fully control the listening environment when the experiment is being conducted. The Sony seem to fit the bill for the job as they have an almost perfect linear response that will be extremely easy to compensate for. As far as my research goes the said model is extremely reliable and should do a great job for the project as well as later listening plus they can be had for under $100.

 

[stratman]

Hey Andrew,

Please post the results of the test, it sounds intriguing.

 

[avaserfi]

Hey Andrew,

Please post the results of the test, it sounds intriguing.

I certainly plan on it. I will likely be starting everything within the month and the project will be due by the end of the semester (mid December) so around that time I will post my full right up here as well. For those who care it will be in full APA format with a complete description of the Rational, methodology, subjects, statistical analysis etc. The idea behind the class is to become familiar with writing journal articles.

My biggest fear currently is getting a solid number of participants, but I should be able to manage. I would be happy with a non-random but diversified sample size of at least 50, but the more the better.

 

[masak_aer]

I would suggest you combine the two suggestions you have in the beginning
of your thread. It will be a lil' bit more complex but will achieve better picture of a person's perception regarding audio compressions in relation to loudness. It would be advisable to use 3 different audio compressions with 3 different level of loudness. I would think it is best to subject the volunteers to listen to the lowest audio compressed then the highest and lastly the mid-audio compressed track. Give them each 3 seconds to rest their ears (i don't know why i pick 3). With simple statistics analysis this can be done pretty easily.

In a research like this, there should be several assumptions need to be made, such as: little or no interference from outside noise, the headphone used has linear frequency response and the audio track covers lowest and highest audible frequency and don't forget, the samples (persons who are picked to be the victims of your research) are not tone deaf...

Just an idea...

PS: May be take a sample of equal number of sexes (male/female). Then you can see how different the two sexes are in this research.