3 months since I started this thread. I have answered my own question. The answer is: move all furniture out of the way. Don't try to correct with a dummy body in situ.
The reason why is that only the minimum phase response of the loudspeaker should be corrected. Everything else is excess phase and should be left alone. For those less well versed in this (such as myself only a few months ago), I will explain.
The measured frequency response at the MLP contains the loudspeaker response (i.e. the minimum phase response) and the room response (i.e. excess phase). The MP response is the only response that can be corrected by inversion, which corrects both amplitude response and phase at the same time. It is also invariable, as in the same loudspeaker will always produce the same MP response. At high freqs, the EP response can simply be windowed out. If a speaker is measured from the MLP, this "invariable" rule holds down to a lower frequency limit - at some point the wavelengths get too long, and early reflections make it impossible to distinguish between what is MP and what is EP.
OTOH, the EP response is highly variable depending on microphone position. This is made even worse if there are nearby reflecting surfaces. This means that performing correction with the EP response, as in having the sofa in situ, only corrects for one point in space and that any movement of the head will render the correction invalid.
This is what I currently think:
1. If the speaker needs correction so that it is flat, correct it for an anechoic flat response at 1m the best that you can. This might mean an outside measurement. After this, leave the upper frequencies alone (Toole's recommendation).
2. If you are unable to perform an anechoic or quasi-anechoic measurement, the next best thing is to do it in-room. This means that the lower frequency limit for "anechoic flat" will be higher, and the limit depends on proximity of reflecting surfaces. The limit can be calculated by speed of sound divided by distance to the closest reflecting surface. Move furniture out of the way and avoid any nearby reflecting surfaces. Make sure the MLP isn't close to a wall, for example.
3. If you are unable to perform 1 or 2, then the next best thing is to perform an MMM or multi-point averaged sweeps incorporating the entire listening area. Some EP will be incorporated into the correction, but it will avoid the problem of correcting for a single point in space.
So what about BACCH's ORC which uses binaural mics which corrects for the HRTF (in addition to sofa, walls, and nearby reflecting surfaces)? I feel that unless the head tracking algorithm is VERY good and can account for head rotation (which will change the HRTF), it has the potential to degrade the sound.
Why you concluded this?
I believe the answer is variable.
If you listen to an infinitesimally small point, you can include everything in the correction.
But it is obviously unreal, we need for a certain volume within which to move. And within that volume there is variability. There is no more or less representative point. Even if you remove all the furniture from the room, because the room still induces variability.
However, in the variability there will necessarily be a significant part that can be corrected (not only the MP). The challenge is to find the right mathematical correlation in each domain (Dirac use the technique of quantifying the deviation from the average response, according to an old white paper, but for me it is not very valid in time domain... Bass Control's new algorithm is much smarter, but it only applies to the lows).
Clearly we do not have AI available, so we fall back on approximations and neglect of something.
But in this way, measuring with furniture, on more than one point, applying appropriate FDW to leave out some patological EP and mediate everything, allows you to obtain information that is neither more nor less acceptable than anything else, because in reality you will never know what result is more or less informative. You could analyze manually.
For example, you can have a high EP for a certain frequency throughout the listening volume, and you would like to correct that... but you have to determine this information manually and set appropriate FDW for that frequency.
In practice, you cannot do this job for the entire frequency range and all the infinitesimal points of the listening volume.
Therefore you will never know how representative your measurement is, especially if on a single point, regardless of whether you have furniture, body or other.
The answer is that environmental factors are a variable of the variables and must be treated as such.
IMO
