Question About Project of Developing an E-Nose for the Detection of Breath VOC Makeup changes

[saanvihandigund]

Hi! We’re building a low-cost e-nose to detect psychological stress from human breath VOCs (Volatile Organic Compounds), using MQ-3, MQ-135, MQ-138, CCS811, and BME688 sensors. The device will feed data to a machine learning model to classify “stressed” vs “relaxed” breath patterns, accounting for temperature and humidity.

How can we improve signal reliability and reduce noise with low-cost gas sensors?
Best bleep to compensate for environmental factors like humidity, temperature, or light?
Tips for building an ML pipeline for multi-sensor, small-sample datasets?

We’re not sure how feasible this project is, but breathomics seemed really interested and we wanted to give it a go! any help is appreciated.