RE: embedding sound problem
RE: embedding sound problem
- Subject: RE: embedding sound problem
- From: "Edwards, Waverly" <email@hidden>
- Date: Wed, 7 Jul 2010 13:31:51 -0500
- Acceptlanguage: en-US
- Thread-topic: embedding sound problem
>> Is your purpose to generate signals for noise-masking perceptual psychology experiments?
Yes, these are for psychology experiments.
>> I wonder what would be 'close enough for rock'n'roll'? Would it be too...
Rock-n-roll would not be used because of the high dynamic range. I would need to be a source that did not vary in a great amount.
>>
If you had some well-defined method of doing it, then it ought to be valid for instance comparing a -25 dB mix against a -20 dB mix or whatever, even if the method isn't exactly 'perfect'? Whatever a perfect method might be.
<<
Unfortunately, I do not have a well-defined method in the digital realm. Using physical objects such as loudspeakers, there is a well-defined way but this research needs to be better controlled and measurable. I have done quite a bit of research on the subject but how is done digitally is not clear to me at this time.
I will work towards using average magnitude but I suspect there is a better way. I just do not know it yet.
Thanks,
W.
-----Original Message-----
From: coreaudio-api-bounces+waverly.edwards=email@hidden [mailto:coreaudio-api-bounces+waverly.edwards=email@hidden] On Behalf Of James Chandler Jr
Sent: Saturday, July 03, 2010 8:15 AM
Cc: email@hidden
Subject: Re: embedding sound problem
Hi Waverly
Is your purpose to generate signals for noise-masking perceptual psychology experiments?
As you say, one might get pretty complicated deciding exacty what a 'precise' level mixing would imply.
I wonder what would be 'close enough for rock'n'roll'? Would it be too imprecise to either normalize both signals, or use dynamic processing to auto-gain both signals, and then do the mixing and assume that the relationship is close enough?
If you use a strong auto-gain dynamics processing on both signals, it ought to run fine in realtime on any two sources, assuming the result is judged 'rigorous' enough.
If you had some well-defined method of doing it, then it ought to be valid for instance comparing a -25 dB mix against a -20 dB mix or whatever, even if the method isn't exactly 'perfect'? Whatever a perfect method might be.
Maybe your purpose is way different than a psych experiment, but researchers were getting 'useful' results in masking experiments with crude mixers and reel to reel tape recorders back in the 1940-1950's.
James Chandler Jr.
On Jul 3, 2010, at 6:38 AM, Edwards, Waverly wrote:
>
> Hello,
>
> I have been working towards embedding/mixing one waveform into another with the end goal of having the carrier be a white noise or music source. The embedded source (speech) amplitude must between -15 to -25 db of the carrier source. Here is the question. What would the best way be to establish this range using an algorithm? I have been trying for days to think of a way that would work.
>
> The only reasonable solution I can think of is to pass through the entire white noise sample obtaining an average magnitude. I would then do the same for the spoken sample samples. Once I've obtain the average of both sample sets, I would then reduce the spoken sample by a percentage so the average does not exceed the value: 10 ^ ( dbLimit /20 ). That seems like a reasonable way to go but if I were using music, during quiet passages the speech would be obvious. I am also concerned there may be too much reduction and there will be a loss of intelligence. May I ask, what you would suggest as a way of achieving this goal?
>
> Ultimately I would like to perform this in realtime but the above described method could only be achieved offline because it requires an average of the squares of the magnitude over the entire sample of both the carrier and embedded source. Maybe I could use a small moving "window" of samples to get the average of squares. I didn't think my goal was going to be a challenge but the more thought I've put into this, the more complex it appears to be. I don't know if the end result would end up being some sort of dynamic automatic gain control, ducking in reverse. All suggestions are welcome.
>
> Thank you for your time,
>
>
> W. _______________________________________________
> Do not post admin requests to the list. They will be ignored.
> Coreaudio-api mailing list (email@hidden)
> Help/Unsubscribe/Update your Subscription:
>
> This email sent to email@hidden
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Coreaudio-api mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Coreaudio-api mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden