Filtering of cross-talk audio signal

I have two microphones (the first placed in the ground floor, the other in the first floor).

In the attached images you can see the recorded audio signals of the two sensors (the second plot shows the RMS value with a window length of 1 second).

My goal is to detect if the source of the sound was next to sensor 1 or next to sensor 2, but only if it was the same sound source (the signals are almost the same, like cross-talk).

So my first approach was, to do a simple cross correlation to detect if the signals are the same and afterwards compare the amplitudes of the signals. But this doesn't give me good results...

Is it possible to solve this issue with an echo cancellation algorithm? Any other ideas?

Normal:

RMS:

[Star Strider]

My first approach would be to see if independent component analysis would work.

[Sivakumaran Chandrasekaran]

sometimes, noise will add and produce different output. we have code for it. to check whether both are same? please reach us at [e-mail] if interested