You do have a point here. The OPN technology is not DNN based, while the More is DNN based, and the Intent is DNN 2.0 based. However, I’ve always presumed that the creation of the DNN is to facilitate neural noise suppression for more speech understanding first and foremost. In other words, it finds a different way to do noise suppression through rebalancing (rather than through subtracting from a noise model like with the OPN).
I always presumed that the DNN was never about improving speech clarity even when neural noise suppression is not applied. After all, the DNN basically just breaks down the sound components in a sound scene and recreates it strictly for the purpose of making feasible the rebalancing of the now discrete sound components individually as an alternative approach for noise suppression. It has never about making all the sound components sound better than before per se even when there’s no noise suppression.
So are you saying that in an easy, quiet environment with no noise suppression activated at all, the sound quality in the More is still actually better than the sound quality in the OPN? I can buy that the sound quality in the Intent with the DNN 2.0 is better than the sound quality in the More with DNN 1.0. But I don’t understand why the DNN 1.0 sound quality would be any better than the OPN sound quality at all. After all, the OPN takes the sound from the input and spits it out “as is” to the output (assuming no noise suppression processing takes place at all in this scenario). So why would a DNN-based recreated sound component have a better sound quality than a direct copy of it from the input to the output (in the OPN case)?
Nevertheless, that’s all just a conjecture using theory. In practice, if you have the OPN and the More and the Intent, and your observation about their sound qualities are good/better/best respectively, then I believe you. I just want to make sure whether the scenario you described precludes any use of any noise suppression in your observation or not. Hence my question above for clarification on whether this observation is encased in the easy/simple environment where noise suppression is not factored in at all.
Using the analogy you started out with, maybe if we take a picture of a visual scene (let’s say it’s a day on the beach), the OPN gives you the whole picture, with EVERYTHING in the background included. But maybe with the More, let’s say if you cut out all the main figures and more visible figures from that picture, like the seagull flying, the tuba player playing in the background, the people walking or playing ball or sun tanning on the beach, the water from the ocean, the sun setting in the backdrop, etc., and you reassemble them all back together into a blank page, those cut-out figures might stand out more because some of the background visuals may be lost or lessened, like the sand of the beach, or the blue sky with some clouds, or the distant mountain in the corner of the bay, etc. Meaning that the cut-out figures now “pops out” more by themselves, hence the perception of clearer quality, compared to the original sound scene displayed by the OPN “as is”.
. The Intents are not perfect though—connectivity with Oticon hearing aids is subpar, and speech in noise is always a challenge—but they are really good.
