I have 12 years experience of audio processing, including 1 patent for audio processing framework (see [login to view URL]).
Your work is easily achieved by transforming the 2 audio inputs into waveforms. The waveforms can be regarded as curves, and the problem is then "a distance" between 2 curves (which has number of ways to solve). I actually did already something quite similar, when I needed to create a module that compare playout audio to set of predefined sound-database.
Depends on the distance between the two sources, the temporal wave form might be transposed to frequency domain.
In any case, all the stages (transpose to wave form, curve, or frequency-domain, and then compare the curves and output the likelihood) will all be native.