You confuse a panda3D, with a spectrograph.
You must first implement it, then think about how to get a sound recording with panda3d.
Theoretically, you can make an asynchronous record from each listener, but think that are not suitable for calculations. Thus, the cycle will switch from one ear to another.
In fact, in network games, not the sound but the position of the sound is transmitted. This is so by the way.
You can alternately change the point (ear) to record sound, you can try to use multi-core, but are not sure of the exact result. Example: multiple audio3d listeners
With this library, you can get data from the microphone, I think, and the application can be written.
docs.python.org/3/library/audioop.html
The book has an example: Lang C. - Panda3D 1.7 Game Developers Cookbook - 2011