Once an audio waveform is decomposed into channels using a filterbank:

f=aud.filterbank(a)

An envelope extraction, for instance, can be computed using this very simple syntax:

e=sig.envelope(f)

Then the channels can be summed back using the sig.sum command:

s=sig.sum(e)

The summation can be divided by the number of channels using the command:

s=sig.sum(…,'Mean')

If we compute for instance an autocorrelation from the envelopes:

ac=sig.autocor(e)

Then we can sum all the autocorrelation using exactly the same sig.sum command:

s=sig.sum(e)

This summation of non-temporal signals across channels is usually called summary.

(From version 0.11 on)

sig.sum(...,'Type',dim) sums along any dimension.