I agree with you @billgem that the 4D sensors technology is kind of an over-the-top hype sensationalism. It’s not really even 4D per se in the first place. The 2 of the 4D have always been there already in just about any major hearing aid brand → the speech detection, and the environment classification. The other 2D are just the 2 motion sensors that detect the up-and-down movement (one of the remaining 2D), and the head turning motion (the other of the remaining 2D). So to call it 4D is overhype sensationalism. And of course 3D is not sexy enough. It has to be 4D to be more sexy. Who knows, in the next generation or two, maybe Oticon will come out with the “Sixth Sense” even, lol.
While the Oticon whitepaper doesn’t reveal how the interpreted intent based on these sensors play out, the Philips whitepaper does (as seen in the screenshot below). And it’s deceptively simple, almost to the point of being rudimentary. It just gets broken into 3 buckets → more SNR help for front speech if no motion (interpreted as intentful listening), default SNR help if occasional motion, and a 360 degrees omni SNR experience with speech if steady motion (like walking). So the interpretations based on the motion, as you can see, is very rudimentary, and the guessed intent can easily be wrong from the user’s real intent. There’s really no magical silver bullet here about knowing the intent of the user, it’s really just a very rough guess which can easily be wrong half of the times.
I think the real advancement of the Intent is the improvement done with their DNN 2.0. By training the DNN 2.0 with more versatile sound scene samples, they’re able to eke out yet another 2 more dB of noise attenuation from the 10 dB that they were able to get out of the Real already. For every 1 dB of SNR improvement, it translates roughly to another 10% of speech clarity. So a 20% of more speech clarity, on top of the already amazing 10 dB of SNR contrast for speech in the Real, is no chump change here.
But of course DNN 2.0 does not sound as sexy as 4D sensor and being able to “read” user intent, and you wouldn’t be able to attach a new catchy name like “Intent” to the new line up, so off we go with the focus on the 4D sensor, while the real hero in my opinion is the more boring DNN 2.0 improvement.