[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: F0 difference limens for unresolved harmonics

Katrin and David,

In an ARO presentation (1995, hopefully published one day), Steve Greenberg
and I used doubly-amplitude-modulated sinusoids in the ~~100-Hz f0 range,
with the carrier-carrier sometimes as high as 10KHz. We also saw that the
pitch DL in these conditions was between 6 and 10 percent, from which we
also inferred that when spectral mediation of pitch is seriously reduced or
altogether eliminated and only temporal cues are left, the DL's are similar
to those found in time discrimination.


At 01:15 PM 4/20/2005, Katrin Krumbholz wrote:
Dear David,

Daniel Pressnitzer, Roy Patterson and I have conducted a series of
experiments some time ago, in which we investigated the lower limit of
pitch (LLP, or LLMP for lower limit of musical pitch) using (i) a melody
task and (ii) a pitch discrimination task (Pressnitzer et al., 2001,
Jasa 109 (5), 2074-2084; Krumbholz et al., 2000, Jasa 108 (3),
1170-1180). In those two papers we show that (i) pitch discrimination
threshold increases from about 1% for pitches above the LLP to about
5-6% for pitches below the LLP (for cosine-phase harmonic tones; the
discrimination threshold below the LLP depends on the nature of the
stimulus). We also show that, contrary to common belief, the LLP does
not actually correpond to the limit of spectral resolvability of
harmonics: In the low-frequency region, the LLP appears to be a TEMPORAL
rather than a SPECTRAL limit. Those two papers also contain a nice
review of the previous literature.

Daniel showed in his paper that the LLP can be modeled by the Meddis and
Hewitt temporal pitch model, and I showed in a paper in the ISH book
from 2001 that, unlike the temporal model, the optimum-processor
spectral pitch net (SPINET) model is unable to explain the LLP for
iterated rippled noises. I can send you pdfs/docs of those papers in a
separate e-mail, if you are interested.

Best wishes from Katrin