Your SE.DSP question is still not clear to me:
- Is there an ordinal variable (e.g. time) to help you with separating latent variables? Or is the problem purely instantaneous?
- How can you quantify "obtain fairly accurate composition"?
- What are your sensor characteristics (esp. regarding saturation)?
Your question, IMHO, is not yet precise enough, or sufficiently formalized to be able to locate it precisely in the Venn diagram of Statistics, Signal Processing, Data science or Computer Science, which overlap to some extend.
Without ordinal variable, I would put it closer to Statistics.