Continuous audio recognition on ESP32 or similar

Hi all ! I am trying to help prevent a recurring crime. ! Any advice on how I could deploy this to an ESP32 or similar, would be great, thanks in advance!
I would like to detect the sound of a metal saw, cutting a lock (padlock, chain etc). We have the sound recorded on CCTV. It is a distinct metal sawing sound, which lasts for a few minutes, until the lock or chain breaks. This is a semi-regular crime for a client, in a public place, which makes it difficult to catch on CCTV. It would be nice to upload the ‘crime’ sounds only, and not the ‘background’ sounds. Crime sounds would be - metal sawing, hammering, and drilling. There are many unpredictable background sounds at this public location - cars driving, doors closing, footsteps, people talking, birds, dogs. If there is a method of only supplying the ‘crime sounds’ only, that would be great. thanks in advance, for any comments or suggestions you have

Download sounds online like on youtube or edit the sound and remove unwanted noise using 3rd party software

1 Like

you can find pure sounds with no background noise online. you just need to choose which one you will use

1 Like

thanks for the suggestions! Do you think I’ll need to also train it for ‘non crime’ sounds (regular park sounds), or can i just train it for the ‘crime sounds’. thanks again