To estimate the P-S separation coefficients I use a different type of objective function. The energy in the P-s and S-p panels is not a suitable measure because we expect to observe converted waves with significant amplitudes.
A human looking at a gather containing P and S waves often uses the moveout of the events to discriminate between then. With this in mind I have chosen to use an objective function that is calculated in constant velocity stack space. If a section contains only upcoming P waves we would expect to see significant amplitudes at P-p and S-p velocities but none at S-s velocities. A section containing only upcoming S-waves should only have energy at S-s and P-s velocities.
To use this separation procedure a set of velocity ``corridors'' must be defined. These corridors define the regions in time/velocity space in which the different event types are expected to occur.
There are several possible objective functions that can be used to estimate the best separation parameters.