This project started out as my final for a UCSD Embedded Course I completed please read my final report paper apologize for any bad grammer. I am going to keep going on this project as I have made some good first steps. I am going to try some new things to determine when the system has found a true audio trigger event that the user needs to be notified about. I am going to try some audio filtering in hardware and maybe software. I am going to get a PCB and custom hardware built, I would like to look into maybe using Tensor Follow for Key Word Recognition or using FFT Audio Analysis.
One stretch goal I'd recommend considering would be a switchable volume limiting element to make the unit even more appropriate for kids. Restricting the output to 85db might eliminate some of the modularity or perhaps there are ways to adapt such a design for various headphone configurations (since different headphone types represent different loads so you could either create a manual or automated detection system). Good luck!