Design and implementation of a 3D sound system
based on the HRTF
Ivana Rossell
Robert Barti
Department d’Acústica, Enginyeria La Salle
[email protected] http://www..salleURL.edu
Abstract
Virtual Reality is one of the hottest buzzwords today in electronics industry. Until recently, the
focus of these efforts was in providing stereoscopic three-dimensional graphics to stimulate our
sense of vision. However, to create believable worlds, visual cues are not sufficient, the field of
auralization, -three dimensional sound- is necessary. There are a number of motivations for
developing auralization systems in addition to the advent of virtual reality. The three dimensional
processing adds an auxiliary creative element to be manipulated by the commercial musician or
record producer, providing a new realm of entertainment to explore. In this paper, a tridimensional
sound system for performing static sound source placement with headphone playback using the
filtering by HRTF is developed.
1 Introduction 2.1 Method
Research on the acoustical and psychological bases of Our approach is based on linear filtering principles.
human sound localization suggests that the primary Let x1(t) represent an electrical signal that drives a
acoustical cues used to determine sound source loudspeaker in free field, and let y1(t) represent the
position are the interaural differences in time of resultant electrical signal from a probe microphone
arrival (∆T) of a sound wave at a listener’s two ears, positioned at a listener’s eardrum. Similary, let x2(t)
the interaural differences in overall intensity (∆I) of represent an electrical signal that drives a headphone,
the sound, and the position-dependent filtering caused with y2(t) the resultant microphone response.
by the interaction of an incoming sound wave with the
folds of the pinnae. In the past quarter century, many The design of the appropiate filter is best described in
studies have shown that cues provided by pinnae the frequency domain.Thus X1 is the Fourier
filtering are more important than previously believed, Transform of the x1 (t) and so forth. The signal
especially for localizing sounds on the median plane coming from the loudspeaker at a specific position at
(where ∆T and ∆I cues are minimized), and for the eardrum can be written as:
establishing the “externalized “ (out of head)
character of sounds in the natural environment.
Y1= X1· Hs·HRTF (1)
In this paper we will describe how we obtain this
position dependent filter called HRTF (Head Related
Transfer Function). This filter will be used, convolved And the signal at the probe microphone from the
with a anechoically signal, to create the impression headphone can be written as:
(auditive event) of receiving a this signal from a
specific direction. Y2= X2· Hh (2)
Where Hs is the loudspeaker transfer function. And
2 The HRTF Hh is the headphone transfer function.
The first phase of our synthesis procedure involves Thus, if x1(t) is equal to x2(t), using both equations we
measurement of the Head Related Transfer Function, obtain the expression that we will give us the HRTF:
HRTF. We can understand this transfer function as
the filter introduced by a subject to the sound HRTF= Y1·Hh/Y2·Hs (3)
incomming to the eardrum in each ear. So there is a
pair of filters for every position of a sound source in
the space.
This transfer function will be measured for a large 4 Receiver model
number of sound source positions with different
subjects. Fist measurements were made on human subjects. In a
second phase of our work we decided to “built” a
3 HRTF measurement head and torso simulator (HATS).
A wideband noise (MLS sequence) was presented We use a comercial dummy and we made an intent to
(either by loadspeaker or headphones) repetitively, make it suitable for our experiments: The face and
and the response at the listener’s eardrum was head of the dummy were covered by latex, the interior
obtained by averaging the output of the probe was filled with absorbent material and the ears were
microphone. perforated to introduce the microphone.
The acoustical response at the eardrum was measured Next pictures show our HATS:
with a condenser microphone (1/2 inch) with a
conical copper cavity coupled to a flexible probe tube
with an outer diameter of 2.5mm. The microphones
used were identical in sensibility and phase response.
The photography shows the system used to collect the
signal.
For free-field measurements, the wideband signal was
transduced by one loudspeaker positioned at 1.5m
from the subject in the anechoic chamber. Next
picture shows the system:
The elevation swept was made by the rotatory axis in
wich the bar was subjected. The azimuthal swept was The receiver model allowed us to obtain standard
made by moving the chair where the subject was sited HRTFs in spite of the individual transfer functions
to make the measurement. The accuracy of the calculated before.
positions was not very good but enough to obtain our
first results. 5 Results and discussion
The HRTFs obtained were satisfactory and provided a
quite good localization performance.
Next figure shows a pair of –40 degree elevation and
135 degree azimuth HRTF mesured.
6 New system
At present, we are finishing a new system to create a
3D Sound space in the anechoic chamber in order to
minimize measurement variability.
In our system we have 19 loudspeakers mounted on a
semicircular arc (3m diameter) the ends of wich were
attached directly above and directly below the
subject. We have a sound source every 10º. The
subject or the receiver model (HATS) is seated on a
rotary and adjustable chair. It has controlled
movements in 10º steps. Thus, we can obtain in
further works all the HRTFs of the whole sphere with
a 10º accuracy.
Nexts photographies show the whole system in the
anechoic room.
Next pictures show a comparison of a 0 degree
azimuth and 72 degree elevation HRTF measured
using different subjects. The localization performance
was better when listening through one’s own HRTF
measurement.
The tracking control, to avoid head movements
during the measurements, is made with a laser beam
The results of filtering these pairs of HRTFs with positioned and fixed on the head of the subject. The
anechoically recorded mono signal and presenting the laser beam has to aim at a target for any azimuth
resulting sound over headphones, create an position.
impression of perceiving the sound from the position
measurement. 7 Conclusions
The accuracy of the auditory event depended on the The system has given satisfactory results and good 3D
nature of the signal to convolve. Broadband signals, sound simulations. The nature of the signal to be
impulsive and complex signals provided more convolved with the HRTF’s has been very important
satisfactory results.
work with human sound localization and to achieve
experiments to examine auditory perceptions
(following Blauert’s steps). We are working too in
designing an auralization system using BRIRs
(Binaural Room Impulse Response) obtained with our
HATS.
References
[1] Frederic L. Wightman i Doris J. Kistler,
“Headphone simulation of free-field listening. I:
Stimulus synthesis”. J. Acoust. Soc. Am. 85 (2),
febrer de 1989.
[2] Frederic L. Wightman i Doris J. Kistler,
“Headphone simulation of free-field listening.
II: Psychophysical validation”. J. Acoust. Soc.
in the impression created when it was presented under Am. 85 (2), febrer de 1989.
headphones. Presenting some signals continuosly has [3] Henrik Moller, Michael Friis Sorensen, Clemen
given movement sensation creating more spectacular Boje Jensen i Dorte Hammershoi, “ Transfer
simulations. The inside the head effect produced by Characteristics of Headphones Measurements
the headphones was a problem observed in the sound on Human Ears”.J. Audio Eng. Soc., Vol 43,
presented to the subjects. Num.4, abril de 1995.
Future lines to develop in further works could be: [4] Henrik Moller, Michael Friis Sorensen, Clemen
Boje Jensen i Dorte Hammershoi, “Head-
- An implementation of a crosstalk canceller to Related Transfer Functions of Human Subjects”.
play the 3D sound trhough speakears. We are J. Audio Eng. Soc., Vol 43, Num.5, maig de
working in this issue just now with the simple 1995.
crosstalk canceller and,after this first step, next [5] Frank Filipanits Jr, “Design and Implementation
will be the implementation of the room effect of an Auralization System with a Spectrum-
canceller. Based Temporal Processing Optimization”.
- Another work we are implementing in our [6] Keith Dana Martin, “A Computational Model of
department is the obtention of the BRIR Spatial Hearing”. Master of Science in
(Binaural Room Impulse Response). We use in Electrical Engineering at the Massatchusets
this project the same HATS we have shown in Institute of Technology.
this paper. [7] Douglas D.Rife i John Vanderkgoy, “Transfer-
function measurement with maximum_length
- As we have mentioned before, we will work with sequences”.J. Audio Eng. Soc., Vol 37, Num.6,
any sound localization experience, in order to juny de 1989.
improve reseach in binaural effects, in human
sound perception. [8] Bill Cardner and Keith Martin, “HRTF
Measurements of a KEMAR Dummy-Head
Microphone”.MITMedia Lab, maig de 1994.
8 Acknowledgments
This work has been the first step to develop a new
workline in our department. We are now finishing our
system to obtain better HRTFs and to experiment in
all kind of binaural experiences. We are going to
