Last week, I visited a client for work to test out a sound switch device. For one reason and another, the kit didn’t pan out on the day (NB: Highly possible it might have been me.. I need to try again). But with the recordings we got, we can now do some fun work and create a melspectogram correlation technique. It might work.. It certainly looks pretty reliable against background noise and talking. You can see our work in progress and try it yourself at github.com/acecentre…
