Session P Tuesday, May 15 13:30 - 18:00 hr Room C/D
Psychoacoustics, Perception and Listening Tests, Part 2
Chair: Sean Olive, Harman International, Northridge, CA, USA
13:30 hr P-1
Detection Threshold
for Tones above 22kHz
Kaoru Ashihara &
Shogo Kiryu
Electrotechnical Laboratory, Tsukuba, Japan
To investigate audibility of ultrasounds contained in a
complex tone, psycho-acoustic experiments were designed. Human subjects were
required to discriminate stimuli with and without components above 22kHz. All
subjects distinguished between sounds with and without ultrasounds only when
the stimulus was presented through a single loudspeaker. When the stimulus was
divided into six bands of frequencies and presented through 6 loudspeakers in
order to reduce intermodulation distortions, no subject could detect any
ultrasounds. It was concluded that addition of ultrasounds might affect sound
impression by means of some non-linear interaction that might occur in the
loudspeakers.
Paper 5401
14:00 hr P-2
Coloration of
Amplitude-Panned Virtual Sources
Ville Pulkki
Helsinki University of Technology, Espoo, Finland
The coloration of amplitude-panned virtual sources is
studied with listening tests and with auditory modeling for anechoic and
reverberant listening. It is found that the amplitude panning produces a
comb-filter effect that is audible in anechoic listening. When the listening
room is reverberant, the effect is less audible depending on the amount of
reverberation. The coloration of a virtual source is dependent on the number of
loudspeakers used to generate it, and it is also dependent on the locations of
loudspeakers.
Paper 5402
14:30 hr P-3
The Use of
Multidimensional Scaling Techniques in Subjective Testing of Sound Quality
Douglas McKinnie &
Francis Rumsey
University of Surrey, Guildford, UK
The need for measures of overall audio quality is
becoming increasingly clear, as is the realization that audio quality is itself
multidimensional in nature. A research design for subjective listening tests is
presented which uses multidimensional scaling of preference judgments. The task
for the listener is greatly simplified in comparison to scaling of multiple
attributes, yet MDS techniques can reveal both the underlying perceptually
relevant dimensions of the stimuli under test and a unidimensional quality
rating in the form of proximity to an ideal point in the model space. Examples
from the authors� research illustrate these points.
Printed Paper not available
15:00 hr P-4
Uses and Misuses of Psychophysical Methods in the
Evaluation of Spatial Sound Reproduction
William Martens
University of Aizu, Aizu-Wakamatsu, Japan
Whereas the primary motivation in spatial hearing research
has been to gain greater understanding of the mechanisms of human spatial
hearing, the motivation for applied research has been the verification and
validation of various spatial audio rendering technologies under development.
This paper outlines some of the uses and misuses of psychophysical methods
typically employed in the subjective evaluation of spatial sound reproduction.
The emphasis is upon the essential tension between engineering goals and
scientific goals, which, while often conflicting, serve to focus psychophysical
research upon resolving disputes between rival theories of how best to simulate
spatial sound fields for a human listener.
Paper 5403
15:30 hr P-5
Recognition of
Everyday Auditory Scenes: Potentials, Latencies and Cues
Vesa Peltonen, Anssi Klapuri, Antti Eronen
& Mikko Parviainen
Tampere University of Technology, Tampere, Finland
A listening test was conducted where the human abilities
in recognizing everyday auditory scenes based on binaural recordings were
studied. The accuracy, latency, and acoustic cues used by the subjects in the
recognition process were analyzed. The average correct recognition rate for 19
subjects was 70% for 25 different scenes, and the average recognition time was
20 seconds. In most cases, the test subjects reported that the recognition was
based on prominent identified sound events.
Paper 5404
16:00 hr P-6
GLS - A Generalized
Listener Selection Procedure
Ville-Veikko Mattila
& Nick Zacharov
Nokia Research Center, Tampere, Finland
The need to perform subjective evaluations of audio
is forever present. Such techniques are known to be inefficient and prone to
unreliability. This can be partially overcome by using so-called expert as
opposed to naive listeners. Expertise is addressed in some depth to clarify its
meaning and to illustrate the benefits in terms of reliability and
repeatability of listening tests. The generalized listener selection (GLS)
procedure is presented for establishing permanent expert listening panels for a
wide range of subjective tests. The method allows for the rapid selection and
assessment of listeners based upon a number of criteria. Correct sampling of
the population is achieved by an assessment of on-line questionnaires, followed
by an audiometric evaluation. The last stage of the GLS procedure consists of
three listening tests, identical in structure, designed to evaluate the
discrimination skills and reliability of subjects. Means for the assessment of
both intra-rater reliability and inter-rater agreement are presented.
Paper 5405
16:30 hr P-7
Perceptual (Ir)relevance of HRTF Magnitude and Phase Spectra
Jeroen Breebaart
(1) & Armin Kohlrausch (2)
IPO Center
for User-System Interaction, Eindhoven, The Netherlands
Philips
Research Laboratories, Eindhoven, The Netherlands
This paper discusses the perceptual consequences of
smoothing of anechoic HRTF phase and magnitude spectra. The smoothing process
is based on a binaural perception model, in which interaural cues in the
auditory system are rendered at a limited spectral resolution. This limited
resolution is the result of the filter bank present in the peripheral auditory
system (i.e. the cochlea). Listening tests with single and multiple virtual
sound sources revealed that both the phase and magnitude spectra of HRTFs can
be smoothed with a gamma-tone filter which equals estimates of the spectral
resolution of the cochlea without audible artifacts. The amount of smoothing
was then increased by decreasing the order of the gamma-tone filters. If the
filter order is reduced by a factor 3, subjects indicate spectral and
positional changes in the virtual sound sources. The binaural detection model
developed by Breebaart, van de Par and Kohlrausch was used to predict the
audibility of the smoothing process. A comparison between model predictions and
experimental data showed that the threshold at which subjects start to hear
smoothing artifacts can be predicted accurately. Moreover, a high correlation
exists between the model output and the amount of stimulus degradation reported
by subjects.
Paper 5406
17:00 hr P-8
Progress in
Perceptual Transfer Function Measurement - Tonal Balance
David Clark
DLC Design, Wixom,
MI, USA
Measured data from over 150 automotive sound systems is
compared to subjective assessment of music tonal balance for the same systems.
The trained listeners assessed a set of sub-attributes of tonal balance, such
as peakiness and frequency extension using music source material. Listening was
always completed before measurements were made. The measured data was analyzed
by a corresponding set of technical sub-attributes. Encouraging correlation was
found between aspects of measured and perceived tonal balance.
Paper 5407
17:30 hr P-9
Development of an
Engineering Application for Subjective Evaluation of Human Response to Noise
Charalampos Dimoulas,
George Kalliris, George Papanikolaou, Dimitris
Christidis & Christos Sevastiadis
Aristotle University of Thessaloniki, Thessaloniki, Greece
This work focuses on the design and implementation of a
computerized psycho-physiological monitoring engineering system that detects
human responses to environmental noises. Classical human stimulus monitoring
analysis tools were used to treat human bodily response under the presence of
annoying noises. Applicable sound recording and reinforcement equipment were
selected and employed to accurately simulate environmental noise conditions. A
computerized data acquisition and analysis system was designed and developed.
Some first results have been obtained and are currently analyzed.
Paper 5408
