All inventions are some novel/non-obvious combination and/or configuration of existing components/methods. Even if you come up with a new chemical element, it is merely a new combination of well-known subatomic particles. Further, even if you discover a new physical force, you can't patent it over 35 USC 101, as it is a "judicially recognized exception".
The point being, there are no truly "new" inventions.
Therefore, the question is "to what degree is this combination distinct from all pre-existing combinations in the solution space".
Practically speaking, you want something distinct enough such that two prior art references in combination do not essentially teach what you disclose. If it takes more than two references, it starts to be a stretch for an examiner/challenger to make an obviousness rejection. The "glue" between the explicit prior art references is the "ordinary skill" of someone who works in this field, so probably someone with an MS/PhD in EE/CS in this case. For example, if the invention was to compress video using neural network auto-encoders with a Fourier transform input, then two research papers individually showing the benefits of auto-encoders and Fourier transforms for efficient lossy compression would probably make that invention obvious, especially since all the sub-components are known in relation to the problem at hand.
Secondary considerations are generally a fairly weak way of proving non-obviousness, especially in front of examiners, who are more likely to stick with the status quo (unlike litigation scenarios, which may offer a more open-minded consideration of the facts). You will at least need an expert affidavit/declaration backing up your secondary consideration argument. Your best secondary consideration arguments would likely be identification of the problem to be solved (which is a core prong of the obviousness inquiry alongside solution to the problem), or teaching away by the prior art (e.g. prior authoritative publication says "this isn't possible for so-and-so reasons", and you prove otherwise).