Re: Autocorrelation

[Christian Kaernbach <chris@xxxxxxxxxxxxxxxxxxxxxxxxxx>]

Dear Bob,

> One thing which struck me is that even the random pulse trains seem to
> have a (weak) pitch.

This is undoubtedly so. We measured the equivalent pitches for quite a
range of random click trains (always high-pass filtered and low-pass
masked). They are often ambigous, but there is a tendency to have a
dominate percept corresponding to the longest first-order interval, i.e.
a kind of edge pitch (if there is an edge). So if, for instance, the
intervals are random and uniformingly distributed from 0 to 10 ms, this
would elicit the same "100 Hz" pitch as if the distribution was from 5
to 10 ms. If the band gets closer, however, an average seems to play a
role (8-10 ms sounds more like 9 ms).

> Indeed, one can successively reduce the pitch of a bandpass filtered
> pulse train (whose harmonics are unresolved by the peripheral auditory
> system) by deleting a greater and greater proportion of those pulses
> (with the deleted pulses being selected at random)

Was the original pulse train a random one?

- Christian