Auditory Scene Analysis

[Thomas savage]

Ah, come on guys, you can have a whip around & do better than that, surely

The spaz jet is being used by ACDC at the moment John, axl loves it.

It's hunt the change, down the back of the sofa time!

 

[Thomas savage]

Well John Only one question remains...,

 

[John Kenny]

Well John Only one question remains...,

Ah, sweet, I didn't know you cared in that way, Tom
Rubbish song, btw but the sentiment is clear - how could it not be with 13 words repeated ad nauseam.

 

[Phelonious Ponk]

Well, Tim, you have taken the listener out of your description & if the listener has a way of analysis (ASA) that is not yet understood & far more sophisticated than current measurements, how do you know what aspects of the signal & correlations between aspects of the signal are of importance? In other words, the current set of measurements are restricted to simplistic signal analysis & yet ASA is a very sophisticated signal analysis.

No, I've taken perception out of the analysis because audio equipment, with the exception of some processors which have been mostly rejected by audiophiles, isn't designed to impact perception, it is designed to reproduce the recording as close as possible, so our perceptions of playback will be as close as possible to the things that were recored. What you're driving at over here is not totally off the wall. After all, stereo itself relies on manipulation of perceptions. Someday some brilliant audio scientist may come up with a whole new recording and reproduction system based on these theories, or maybe even invent a processor that works much better than, oh I dunno...the Aphex Aural Exciter...? That will be really interesting when it happens. In the meantime, I don't think it's accurate to characterize disinterest in this subject as anything more than what it is: People gravitating toward discussion that is relevant to audio reproduction as it exists today.

Tim

 

[Sal1950]

Can you stretch to a taxi(cab) or are we putting John on public transport?

Whatever it takes Thomas, well worth the cost.

 

[John Kenny]

No, I've taken perception out of the analysis because audio equipment, with the exception of some processors which have been mostly rejected by audiophiles, isn't designed to impact perception, it is designed to reproduce the recording as close as possible, so our perceptions of playback will be as close as possible to the things that were recored.

It's an unfortunate fact that almost all of us judge the result of that goal by listening to it & not by reading measurements. Maybe you're different, Tim but last time I checked you still listen to music playback as do the rest of us who are engaged in this hobby. If the hobby was who can build the best measuring device then your point would be valid but we are left to the vagaries of our auditory perception & it can only benefit us to better understand how this perception works.

What you're driving at over here is not totally off the wall. After all, stereo itself relies on manipulation of perceptions.

Aha! You get some of the concept then, Tim, - the whole audio reproduction chain is an attempt at creating an illusion which is believable to the LISTENER & this is where auditory perception comes into it, Tim

Someday some brilliant audio scientist may come up with a whole new recording and reproduction system based on these theories, or maybe even invent a processor that works much better than, oh I dunno...the Aphex Aural Exciter...? That will be really interesting when it happens.

Maybe or maybe a better understanding of exactly what our auditory perception considers important to realism will help to produce a more believable playback?

In the meantime, I don't think it's accurate to characterize disinterest in this subject as anything more than what it is: People gravitating toward discussion that is relevant to audio reproduction as it exists today.

Tim

Well, forgive me for reminding people who cynically dismiss audio devices that there may actually be something of perceptual value in these devices

 

[Phelonious Ponk]

Well, forgive me for reminding people who cynically dismiss audio devices that there may actually be something of perceptual value in these devices

Someone is cynically dismissing audio devices? I thought this was about ASA and didn't realize it had even been related to any audio devices.

Tim

 

[John Kenny]

Someone is cynically dismissing audio devices? I thought this was about ASA and didn't realize it had even been related to any audio devices.

Tim

You can have the last word, Tim

 

[fas42]

Hello John

Any peer reviews? Any AES references to the works of Bergman? Some metrics, measurements? Anything more substantive?

Hello, Frantz ... the work of Bregman has kicked off an enormous amount of research into the concepts he espoused, but I doubt AES mentions it at all - it has all happened in neuroscience, psychophysics. It was a revelation to me, because it explained the seemingly bizarre behaviour of audio suddenly snapping into a higher level of perceived quality when the level of audible artifacts dropped below a certain point. This was something that had intrigued me for decades, I knew that it happened but I didn't know why, from the POV of how my head worked.

The relevance to audio reproduction is pretty obvious, IMO - the brain will do a lot of work of 'understanding' what the sound waves impinging on it mean, but too many artifacts disrupt the interpretation of what is being heard. In the jargon of that field of research, the number of Mismatch Negatives (MMN) is too great - "The mismatch negativity (MMN) is a brain response to violations of a rule" - the rule that the brain may be following is, for example, "the sounds I'm hearing are the voice of a real person". Therefore, the aim of the audio system should be to prevent MMNs happening - and this is what I managed to make occur, unintentionally, 30 years ago ...

 

[Don Hills]

... too many artifacts disrupt the interpretation of what is being heard. ... Therefore, the aim of the audio system should be to prevent MMNs happening ...

... and we can already measure the effectiveness of any attempts to reduce the artifacts ("MMNs"). We don't have to listen. What does need further research is the effect on ASA of the common artifacts. For example, we can quantify a change in noise level caused by a reproduction system, but we need to do more to quantify its effect on ASA. Most of the work to date has been on quantifying the audibility of various artifacts - for example, the audibility of signal-related noise modulation in the presence of masking.

 

[Phelonious Ponk]

You can have the last word, Tim

The problem being that the last word was a serious question. I am actually aware of some processors that work on perception - hall simulators in AVRs, headphone processors that emulate surround, etc. Most audiophiles consider them cheap tricks, but I've heard a couple of headphone processors that simply try to get a bigger space, get rid of that "inside your head" feeling, that were pretty impressive. Don't know if I'd get tired of them after a few hours of listening or not. Are there others that you're aware of? Any of them based on learnings from ASA? That would be interesting to explore.

Tim

 

[fas42]

... and we can already measure the effectiveness of any attempts to reduce the artifacts ("MMNs"). We don't have to listen. What does need further research is the effect on ASA of the common artifacts. For example, we can quantify a change in noise level caused by a reproduction system, but we need to do more to quantify its effect on ASA. Most of the work to date has been on quantifying the audibility of various artifacts - for example, the audibility of signal-related noise modulation in the presence of masking.

Not in my experience. Just two examples I have had to deal with with is the impact of poor connections in the system, and various interference mechanisms. I have not seen any research quantifying these factors, yet IME these are crucial, must be dealt with - I don't hear these artifacts directly, what I hear is sound that is "not right" - I'm aware of the impact indirectly.

This is the value, for me, of the work I've done over the years - learning to adjust the level of these degrading influences, and hearing how that impacts on the perceived sound. It's very apparent that no-one is actually measuring any of this, mostly because they don't know how to.

 

[AJ Soundfield]

Another interesting resource on ASA, and other sound perception matters: http://home.ieis.tue.nl/dhermes/lectures/soundperception/07ASA1.html. Has a nice summary of Bregman's concepts, and points to related areas of research, etc.

http://home.ieis.tue.nl/dhermes/lectures/soundperception/01Introduction.html

This lecture series aims at discussing the problems which underlie these issues, and to explore some ways where we can find a solution to these problems. The emphasis is on the perception and the design of non-speech and non-music sounds. The synthesis of musical sounds and speech sounds is widely dealt with elsewhere, and the state of the art in these research fields can be found in many good textbooks.

What does any of this have to with audiophile massive self delusion and high susceptibility to suggestion?

 

[NorthSky]

https://auditoryneuroscience.com/scene_analysis

 

[fas42]

This lecture series aims at discussing the problems which underlie these issues, and to explore some ways where we can find a solution to these problems. The emphasis is on the perception and the design of non-speech and non-music sounds. The synthesis of musical sounds and speech sounds is widely dealt with elsewhere, and the state of the art in these research fields can be found in many good textbooks.

What does any of this have to with audiophile massive self delusion and high susceptibility to suggestion?

Because the intent of the lectures is to deal with

In this lecture series we will discuss various fundamental and applied issues which arise in designing the sounds for complex systems and for virtual environments. These issues arise from the observation that in most computer-controlled systems the use of sound is either limited to short attention signals consisting merely of simple tone beeps, or is completely absent

So the aim is for the students to understand how to create sounds for use in drawing people's attention to something - which they can do a better job of if they know how human hearings works in understanding the environment - which at times would include music, which is our interest.

Yes, our hearing does in fact delude us, and make us susceptible to believing something is real when it's not - but this is something we want to make use of, very deliberately!! The fact that people get convinced that some silly object, or technique is making things sound more real is just part of the total picture of what's going on here - if we do this the right way, then the audio sounds more real, all the time - even when we lose interest in that enthusiasm, it will still do the job of producing better sound, in terms of how we hear it.

 

[fas42]

John, elsewhere, pointed to this paper - I don't think it's been mentioned here yet: http://audition.ens.fr/dp/pdfs/pressnitzer-2011-music_ambiguity_ASA.pdf - "Auditory scene analysis: the sweet music of ambiguity". Nice layman's look at the power of illusion, and how it's actually useful to us humans to be fooled! Key terms are "ambiguity" and "ecological plausibility" - if we experience something, and our senses give us enough information to interpret the sensation in more than one way, then our "animal" minds will pick what is "more likely", even if our "intellectual" minds "know" the real story ...

In the hifi context, normally the data tells us that the sound is 'fake', every time ... there is zero ambiguity here! But at elevated levels of quality the distinction starts to become blurred - from my POV, the disappearing speaker illusion occurs because our "animal" minds have decided that the ecologically most plausible explanation is that the projected auditory scene is 'real', even though in every other sense we know that it can't be so.

 

[Bill Bainbridge]

Frank, thanks for the tip to check this thread out. Quite excellent links, folks and thank you!
There's so much to digest it will take some time for me to reflect adequately regarding direct application to our common listening experience, but there is one item that this might yield: an improved ABX (device and software) which recorded all manipulations of the auditioning switches to provide data regarding user timing of selections, and a feature which would allow manipulations of envelope between source switching (a sort of ADSR) that would also be capable of programming those changes into an event step-sequencer. With some careful set-up using parameters informed from ASA study, such a device might overcome some of the issues with current fixed (and abrupt) envelope changes that may still encumber the ABX process.

I'll have to review the papers again to be sure of the source, but one of the research heavyweights (maybe Bregman?) suggested that a combination of objective and subjective testing would be the best to remove resultant problems with possible "perceptual lensing" (my paraphrase-term) from the final numerical results.

Frank, I'm curious if your preferred playback system included Dipole Drivers in the speaker set? I'm also curious to know if your rejection of acoustic treatments included those only focused on first reflection points, and why?

 

[fas42]

Bill, no, conventional box speakers at all times. I don't actually reject acoustic treatments, just that I don't consciously include them - I listen in typical somewhat messy rooms, no special place is set up for a system ... and I have never felt the need for worrying about them. I have been in listening rooms that were very heavily treated, and I found them disturbing, unsettling - I could not enjoy listening to music in that type of space.

 

[Bill Bainbridge]

I have been in listening rooms that were very heavily treated, and I found them disturbing, unsettling - I could not enjoy listening to music in that type of space.

That's an understandable response to the common ways in which acoustic absorption (particularly) is misapplied. It has taken a few decades for recording studio control rooms (used for mixing and mastering) to develop better criteria for choosing a room, and smoothing its response, such as contained in:
http://www.itu.int/rec/R-REC-BS.1116

Of note there has been an amazing amount of information offered online in the past few years by some very talented and gracious acousticians, such as John Sayers, Andre Vare, Eric Desart, to name just a few. With their technical insight and the help of practical builders such as Rod Gervais, and acoustic product developers such as Ethan Winer (Realtraps), Glen Kuras (GIK), and of course Floyd Toole (RPG) its possible for a person of modest means to develop an excellent space for the production and refinement of audio.
I've been working on what will become a modest rehearsal/composition/recording/mixing/mastering room for nearly six years now- designing, building, outfitting, and am currently involved in the acoustic treatment design and execution phase for it. I had to complete it all DIY-style due to budgetary concerns, but I can assure you practical acoustics are not a trivial exercise to get right.

Having said that, I can appreciate those who prefer their listening environment to be less specialized. I do believe that some of that acoustical technology could benefit those who would seek to have their room be less of a negative influence on their system response, however.

 

[fas42]

Bill, I appreciate that the recording environment is a completely different kettle of fish - here the acoustics are part of the production of the sound, as a creative process. And I can imagine that people who spend a lot of time working in such places develop a certain "ear", in terms of what they listen for - my personal interest is in having a replay system "fool" one, a blindfolded person would not pick the deception; and in my journey I've discovered that the electronics aspects of the listening chain are the most critical, for getting this to happen.