I want to be able to store some sampled sounds, compare them to live samples and trigger an event when they are the same ...
Example - Save the sound of a footstep being made on a gravel path (in fact several at different attenuations) and detect and trigger when a fresh footstep is heard. A fairly simple security application but hopefully reasonably foolproof. It has to be a tiny self contained circuit that can be reproduced cheaply - obviously dead easy to do with a PC and a speech recognition package but how do I do it all in a tiny module?
Any ideas on how to do that please or pointers towards anything that does that job that pre exists?
I think it all depends on how reliably the device should be in terms of false positive/false negative. If you can afford low reliability, you can get away with a few bandpass filters or FFT implemented on a primitive DSP processor and compare spectrums. If the reliability is important, you will neeed a fairly powerful DSP processor doing something like "dynamic programming" (an algorithm for comparing patterns).
Thanks for that, I guess (as always) it's down to cost. I have now found a cheap sound activated 'glass break' detector ($15) will take one apart and see if I can adjust the filtering to make it react to other frequencies.