From: malcolm@interval.com (Malcolm Slaney)
Date: Thu, 11 Jan 1996 20:21:32 -0800
Subject: Bev Wright and Detecting Unexpected Sounds
Message-Id: <v02130502ad1b8be7de75@[199.170.108.19]>

The Hearing Seminar returns to Stanford next Thursday.  Hurray.  I've got a
pretty good set of volunteers lined up to give talks and discuss their
work.  There are still plenty of empty slots, so let me know if you have
ideas.

This coming Thursday, Bev Wright, a psycoacoustician at UC San Francisco,
will be here to discuss her work with detecting unexpected tones.  This
week we return to the problem of expectations and attention.  How do our
expectations affect our ability to detect new sounds.  More importantly,
for this talk, how do our expectations affect our ability to hear sounds
with novel temporal properties (like different lengths)?

There are applications of this work to compression and even music (not to
mention helping us to understand audition).  Bev will talk about her tests
and the methodology. (The applications are up to you attendees.)

Dr. Wright is also the recent winner of a prestigous award from the
Acoustical Society of America.  Bev won the R. Bruce Lindsay award, which
is given to a member of the ASA under the age of 35 for "contributions to
the understanding of auditory processing of complex signals." From the Dec
'95 ASA p. 3005.  Congrats!

        Who:    Bev Wright (UCSF)
        What:   Detection of sounds having unexpected temporal properties
        When:   11AM on Thursday January 18th
        Where:  CCRMA Library (Top Floor of the Knoll at Stanford)

I've attached the current (working schedule) to the end of this note.  We
do have a number of very interesting talks lined up and you'll want to
leave room on your calendars!

See you at CCRMA.

-- Malcolm

*********************** THE HEARING SEMINAR SCHEDULE ************************

1/11    Jack Loomis (UCSB) Externalization of Virtual Sound **** at SKI
1/18    Bev Wright (UCSF) Detection of Unexpected Sounds
1/25    Nelson Morgan (ICSI) SPAM
2/1     Beth Wenzel (NASA) Tentative!!!
2/8     No Seminar (ARO Meeting)
2/9     Hiroshi Riquimaroux (Doshisha University) Friday Seminar Tentative ***
2/15    Ted Lewis (Berkeley)  4:30PM seminar!
2/22    Chas Pavlovic (ReSound) Hearing Aids
3/7     John Tennison (SU-Med) Modeling Tonal Hierarchy Research
3/14
3/21
3/28    No Seminar
4/4     No Seminar (Acta Acoustica meeting on auditory models)
4/11    Kristin Precoda (EE) Confusability of Speech Sounds
4/18
4/25
5/2
5/9     No Seminar (ICASSP Conference)
5/16
5/23
5/30
6/6     Durand Begault (NASA) Virtual Early Echo Thresholds
6/13
6/20




To Schedule (already accepted)
        Steve Greenberg (UCB) March
        Christof Schreiner (UCSF) Spectral Ripple after ARO
        Sue McCabe (Univ. Plymouth) Auditory Scene Analysis 1/half September
                suem@soc.plym.ac.uk
        Dan Ellis (MIT/ICSI) Sound Separation (After mid May)
        William Banks (Pomona) Negative Auditory Priming  Early May
                                (wbanks@pomona.claremont.edu)
        Duda Student (SJSU) End of semester
        Malcolm Slaney (Interval) After ICASSP



********************* BEV WRIGHTS (1/18) ABSTRACT ***************************

Detection of sounds having unexpected temporal properties.
Bev Wright (UCSF)

        Nearly 30 years ago, Greenberg and Larkin first reported that a
tone in continuous noise is better detected at an expected than an
unexpected FREQUEUNCY.  In their experiment, expectation was manipulated
using the probe-signal method.  That method leads the listener to expect
a target frequency by presenting the signal most often at the target
frequency, and only occasionally at other unexpected probe frequencies.
The resulting function relating percent correct and signal frequency
indicates the detectability of signals at both the expected and
unexpected frequencies.

        Recently, Huanping Dai and I observed a similar result in the
temporal domain.  In our initial experiments, we also employed the
probe-signal method, but presented a signal of fixed frequency at
expected and unexpected DURATIONS.  When each signal duration was
presented alone and expected, it was detected on about 85% of the
trials.  However, when the listeners were lead to expect a short- or a
long-duration signal, signal detectability decreased to near chance as
the signal duration deviated from the expected value.  This occurred when
the signal was a tone at 250, 1000, or 4000 Hz, or a braodband noise.  In
contrast, when lead to expect a range of durations, detectability was
only slightly worse than when each signal was presented alone.  These
results suggest that listeners are capable of adjusting their
temporal-integration interval according to the demands of the task.

        In our two newest experiments, we have seen near chance
detectabilty for (1) unexpected rates of sinusoidal amplitude
modulation--an outcome contrary to the predictions of several popular
models, and (2) brief tonal signals presented at unexpected times after a
noise forward masker--indicating the temporal precision of focused
listening following a clear auditory temporal marker.

        Taken together, these experiments show that the probe-signal
method is an extremely useful technique for determining the detection
cues used by listeners in both the frequency and temporal domains.